Nothing
tests/manual/test.mutator-oncoSimulSample-long.R
In OncoSimulR: Forward Genetic Simulation of Cancer Progression with Epistasis
## Repeat tests in test.mutator, using oncoSimulSample.
## This is a concession to extreme paranoia.
cat(paste("\n Starting test.mutator-oncoSimulSample-long.R test at", date()))
cat(paste("\n a runif ", runif(1), "\n"))
## RNGkind("L'Ecuyer-CMRG") ## for the mclapplies
## require(car) ## for linearHypothesis, below. In the namespace
enom <- function(name, mu, ni = no, pp = pops) {
## expected without a given name for init
ii <- which(names(mu) == name)
out <- ni * pp * mu[-ii]
out[order(names(out))]
}
pnom <- function(name, mu, ni = no, pp = pops) {
ee <- enom(name, mu, ni, pp)
ee/sum(ee)
}
snomSampl <- function(name, out) {
## observed without the init
cs <- colSums(out$popSample)
ii <- which(names(cs) == name)
cs <- cs[-ii]
cs[order(names(cs))]
}
smSampl <- function(name, out) {
## totals for a given gene
cs <- colSums(out$popSample)
ii <- which(names(cs) == name)
cs[ii]
}
totalindSampl <- function(out) {
## total num indivs
sum(out$popSummary$TotalPopSize)
}
medianNClonesOSS <- function(x) {
median(x$popSummary[, "NumClones"])
}
NClonesOSS <- function(x) {
x$popSummary[, "NumClones"]
}
## ugly hack. Of course, not really mutations per clone. But the closest iwth oncoSimulSample and sampling whole pop.
mutsPerCloneOSS <- function(out) {
rowSums(out$popSample)
}
p.value.threshold <- 0.005
## very slow, because huge number of clones. But tests several phenomena comprehensively.
## same with McFL below
date()
test_that("per-gene-mut rates and mutator", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n oss11-ossl: a runif is", runif(1), "\n")
ng <- 40
ni <- rep(0, ng)
m1 <- runif(ng, min = 1e-7, max = 5e-6)
m2 <- rep(1e-5, ng) ## runif(ng, min = 1e-5, max = 1e-4): ## crazy num of clones
names(ni) <- names(m1) <- names(m2) <- c(replicate(ng,
paste(sample(letters, 12), collapse = "")))
fe1 <- allFitnessEffects(noIntGenes = ni)
ft <- 25 ## 50 this is crazy and takes forever
no <- 5e5
reps <- 40
gn <- paste(names(ni), collapse = ", ")
mutator1 <- allMutatorEffects(epistasis = c("MU" = 20),
geneToModule = c("MU" = gn))
mutator2 <- allMutatorEffects(epistasis = c("MU" = 40),
geneToModule = c("MU" = gn))
m1.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m1.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m1.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
runif(1)
m2.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m2.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m2.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
if(! (
inherits(m1.mutator0$popSummary, "data.frame") &&
inherits(m1.mutator1$popSummary, "data.frame") &&
inherits(m1.mutator2$popSummary, "data.frame") &&
inherits(m2.mutator0$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") ) ) {
T8 <- FALSE
cat("\n not a data frame?\n")
}
if(T8) {
m1.mutator0$popSummary[, c(1:3, 8:9)]
m1.mutator1$popSummary[, c(1:3, 8:9)]
m1.mutator2$popSummary[, c(1:3, 8:9)]
m2.mutator0$popSummary[, c(1:3, 8:9)]
m2.mutator1$popSummary[, c(1:3, 8:9)]
m2.mutator2$popSummary[, c(1:3, 8:9)]
## Mutator increases if larger mutator and compared to no mutator
## within levels of per-gene mutation rates
T1 <- ( wilcox.test(m1.mutator2$popSummary[, "NumClones"], m1.mutator1$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T2 <- ( wilcox.test(m1.mutator1$popSummary[, "NumClones"], m1.mutator0$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T3 <- ( wilcox.test(m2.mutator2$popSummary[, "NumClones"], m2.mutator1$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T4 <- ( wilcox.test(m2.mutator1$popSummary[, "NumClones"], m2.mutator0$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
## Increases in mutation rates increase clones, etc, within levels of
## mutator.
T5 <- ( wilcox.test(m2.mutator0$popSummary[, "NumClones"], m1.mutator0$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T6 <- ( wilcox.test(m2.mutator1$popSummary[, "NumClones"], m1.mutator1$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T7 <- ( wilcox.test(m2.mutator2$popSummary[, "NumClones"], m1.mutator2$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
## expect_true( mean(mutsPerCloneOSS(m1.mutator0)) <
## mean(mutsPerCloneOSS(m2.mutator0)))
## expect_true( mean(mutsPerCloneOSS(m1.mutator1)) <
## mean(mutsPerCloneOSS(m2.mutator1)))
## expect_true( mean(mutsPerCloneOSS(m1.mutator2)) <
## mean(mutsPerCloneOSS(m2.mutator2)))
}
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL: per-gene-mut rates and mutator", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mcfloss11-ossl: a runif is", runif(1), "\n")
ng <- 40
ni <- rep(0, ng)
m1 <- rep(5e-6, ng) ## too much variation and hard to pick the diffs.;
## runif(ng, min = 1e-7, max = 5e-6) And if too
## tiny, you do not pick them up unless huge ft
## and then it is way too slow for m2, etc.
m2 <- rep(5e-5, ng)
names(ni) <- names(m1) <- names(m2) <- c(replicate(ng,
paste(sample(letters, 12), collapse = "")))
fe1 <- allFitnessEffects(noIntGenes = ni)
ft <- 20 ## 50
no <- 5e5
reps <- 20 ## 40
gn <- paste(names(ni), collapse = ", ")
mutator1 <- allMutatorEffects(epistasis = c("MU" = 20),
geneToModule = c("MU" = gn))
mutator2 <- allMutatorEffects(epistasis = c("MU" = 40),
geneToModule = c("MU" = gn))
m1.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m1.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m1.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
cat("\n starting m2\n")
m2.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m2.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m2.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
if(! (
inherits(m1.mutator0$popSummary, "data.frame") &&
inherits(m1.mutator1$popSummary, "data.frame") &&
inherits(m1.mutator2$popSummary, "data.frame") &&
inherits(m2.mutator0$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") ) ) {
cat("\n not a data frame?\n")
T8 <- FALSE
}
if(T8) {
m1.mutator0$popSummary[, c(1:3, 8:9)]
m1.mutator1$popSummary[, c(1:3, 8:9)]
m1.mutator2$popSummary[, c(1:3, 8:9)]
m2.mutator0$popSummary[, c(1:3, 8:9)]
m2.mutator1$popSummary[, c(1:3, 8:9)]
m2.mutator2$popSummary[, c(1:3, 8:9)]
## Mutator increases if larger mutator and compared to no mutator
## within levels of per-gene mutation rates
## we could use wilcoxon or t tests actually, specially because often diffs
## are not huge.
p.fail <- 0.005
T1 <- ( t.test( m1.mutator2$popSummary[, "NumClones"] ,
m1.mutator1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( m1.mutator1$popSummary[, "NumClones"] ,
m1.mutator0$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T3 <- ( t.test( m2.mutator2$popSummary[, "NumClones"] ,
m2.mutator1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T4 <- ( t.test( m2.mutator1$popSummary[, "NumClones"] ,
m2.mutator0$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
## Increases in mutation rates increase clones, etc, within levels of
## mutator.
T5 <- ( t.test( m1.mutator0$popSummary[, "NumClones"] ,
m2.mutator0$popSummary[, "NumClones"], alternative = "less")$p.value < p.fail)
T6 <- ( t.test( m1.mutator1$popSummary[, "NumClones"] ,
m2.mutator1$popSummary[, "NumClones"], alternative = "less")$p.value < p.fail)
T7 <- ( t.test( m1.mutator2$popSummary[, "NumClones"] ,
m2.mutator2$popSummary[, "NumClones"], alternative = "less")$p.value < p.fail)
}
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Mutator increases by given factor with per-gene-mut rates: major axis and chi-sq test", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Two cases: mutator and no mutator, with variable mutation rates.
## rates such that rates of no mutator = rates of mutator * mutator.
## Why not compare mutlitplication factor keeping mutation rates
## constant? Because specially with mutator and large diffs in mut
## rates, with oncoSimulSample you undersample variation with
## wholePop, etc.
## Setings similar to oss11 in per-gene-mutation-rates but with the mutator
cat("\n AEu8_long-ossl: a runif is", runif(1), "\n")
pops <- 8000
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg1, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg1.b))
## Compare with the expected for this scenario
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
pg2 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3)
names(pg2) <- names(pg1)
m1.pg2.b <- oncoSimulSample(pops,
fe,
mu = pg2,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg2.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg2.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg2.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg2.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg2, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg2.b))
p.fail <- 1e-3
T2 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg2.b),
p = pnom("oreoisasabgene", pg2, no, pops))$p.value > p.fail)
## Compare the mutator with the no mutator
T3 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
snomSampl("oreoisasabgene", m1.pg2.b))$p.value > p.fail)
y <- sqrt(snomSampl("oreoisasabgene", m1.pg1.b))
x <- sqrt(snomSampl("oreoisasabgene", m1.pg2.b))
mma <- smatr::ma(y ~ x, slope.test = 1, elev.test = 0) ## From smatr package, for major axis
## intercept not different from 0
T4 <- (mma$elevtest[[1]]$p > p.fail)
T5 <- (mma$slopetest[[1]]$p > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL: Mutator increases by given factor with per-gene-mut rates: major axis and chi-sq test", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Two cases: mutator and no mutator, with variable mutation rates.
## rates such that rates of no mutator = rates of mutator * mutator.
## Why not compare mutlitplication factor keeping mutation rates
## constant? Because specially with mutator and large diffs in mut
## rates, with oncoSimulSample you undersample variation with
## wholePop, etc.
## Setings similar to oss11 in per-gene-mutation-rates but with the mutator
cat("\n MCFL: long_AEu8-ossl: a runif is", runif(1), "\n")
pops <- 8000
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
model = "McFL",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg1, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg1.b))
## Compare with the expected for this scenario
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
pg2 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3)
names(pg2) <- names(pg1)
m1.pg2.b <- oncoSimulSample(pops,
fe,
mu = pg2,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
model = "McFL",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg2.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg2.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg2.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg2.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg2, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg2.b))
p.fail <- 1e-3
T2 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg2.b),
p = pnom("oreoisasabgene", pg2, no, pops))$p.value > p.fail)
## Compare mutator with no mutator
T3 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
snomSampl("oreoisasabgene", m1.pg2.b))$p.value > p.fail)
y <- sqrt(snomSampl("oreoisasabgene", m1.pg1.b))
x <- sqrt(snomSampl("oreoisasabgene", m1.pg2.b))
mma <- smatr::ma(y ~ x, slope.test = 1, elev.test = 0) ## From smatr package, for major axis
## intercept not different from 0
T4 <- (mma$elevtest[[1]]$p > p.fail)
T5 <- (mma$slopetest[[1]]$p > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Mutator, several modules differences", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mmd1_2-ossl: a runif is", runif(1), "\n")
reps <- 140
no <- 5e3
ft <- 80 ## you need it large enough to get enough hits
mu <- 1e-5
ln <- 50
m1 <- 7 ## if this is too large, easy to get it to blow.
ni <- rep(0, 2 * ln)
gna <- paste0("a", 1:ln)
gnb <- paste0("b", 1:ln)
names(ni) <- c(gna, gnb)
gn1 <- paste(c(gna, gnb), collapse = ", ")
gna <- paste(gna, collapse = ", ")
gnb <- paste(gnb, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn1))
mut2 <- allMutatorEffects(epistasis = c("A" = m1,
"B" = m1),
geneToModule = c("A" = gna,
"B" = gnb))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
## b2$popSummary[, c(1:3, 8:9)]
## b1$popSummary[, c(1:3, 8:9)]
## mean(rowSums(b2$popSample))
## mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
## Variances for NumClones are hugely unequal, even after log transform.;
## might want Wilcoxon? Similar for rowSums of popSample
p.fail <- 0.05
T1 <- ( wilcox.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( wilcox.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
## Remember that numClones is underestimated, possibly severly, by
## oncoSimulSample compared to oncoSimulPop, since we only look at the
## clones that exist at the end.
date()
test_that("Mutator, several modules differences, McFL", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mmd1_2_mc-ossl: a runif is", runif(1), "\n")
reps <- 80
no <- 5e3
ft <- 50 ## you need it large enough to get enough hits
mu <- 1e-5
ln <- 50
m1 <- 7 ## if this is too large, easy to get it to blow.
ni <- rep(0, 2 * ln)
gna <- paste0("a", 1:ln)
gnb <- paste0("b", 1:ln)
names(ni) <- c(gna, gnb)
gn1 <- paste(c(gna, gnb), collapse = ", ")
gna <- paste(gna, collapse = ", ")
gnb <- paste(gnb, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn1))
mut2 <- allMutatorEffects(epistasis = c("A" = m1,
"B" = m1),
geneToModule = c("A" = gna,
"B" = gnb))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, model = "McFL"
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, model = "McFL"
)
gc()
b2$popSummary[, c(1:3, 8:9)]
b1$popSummary[, c(1:3, 8:9)]
mean(rowSums(b2$popSample))
mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
p.fail <- 0.05
T1 <- ( t.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
## Slow (~ 3 seconds) but tests modules of mutator nicely.
date() ## Beware: this uses a lot of RAM without the gc()
test_that("Mutator modules differences", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mmd1-ossl: a runif is", runif(1), "\n")
reps <- 40
no <- 5e3
ft <- 100
mu <- 1e-5
lni <- 50
m1 <- 1
m2 <- 25
m3 <- 50
ni <- rep(0, lni)
gn <- paste0("a", 1:lni)
names(ni) <- gn
gn <- paste(gn, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn))
mut2 <- allMutatorEffects(epistasis = c("A" = m2),
geneToModule = c("A" = gn))
mut3 <- allMutatorEffects(epistasis = c("A" = m3),
geneToModule = c("A" = gn))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b3 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut3,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b3$popSummary[, c(1:3, 8:9)]
b2$popSummary[, c(1:3, 8:9)]
b1$popSummary[, c(1:3, 8:9)]
## mean(rowSums(b3$popSample))
## mean(rowSums(b2$popSample))
## mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
p.fail <- 0.05
T1 <- ( t.test( b3$popSummary[, "NumClones"],
b2$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T3 <- ( t.test( rowSums(b3$popSample) ,
rowSums(b2$popSample), alternative = "greater")$p.value < p.fail)
T4 <- ( t.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Relative ordering of number of clones with mutator effects", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n x1-ossl: a runif is", runif(1), "\n")
pops <- 40
fe <- allFitnessEffects(noIntGenes = c("a" = 0.12,
"b" = 0.14,
"c" = 0.16,
"d" = 0.11))
fm6 <- allMutatorEffects(noIntGenes = c("a" = 5,
"b" = 5,
"c" = 5,
"d" = 5))
nc1 <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =250,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e6,
detectionSize = 1e8,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm8 <- allMutatorEffects(noIntGenes = c("a" = 1,
"b" = 1,
"c" = 1,
"d" = 1))
nc2 <- oncoSimulSample(pops, fe, muEF = fm8, finalTime =250,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e6,
detectionSize = 1e8,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm7 <- allMutatorEffects(noIntGenes = c("a" = 1e-3,
"b" = 1e-3,
"c" = 1e-3,
"d" = 1e-3))
nc3 <- oncoSimulSample(pops, fe, muEF = fm7, finalTime =250,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e6,
detectionSize = 1e8,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
T1 <- (wilcox.test(nc1$popSummary[, "NumClones"], nc2$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T2 <- (wilcox.test(nc2$popSummary[, "NumClones"], nc3$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
## rowSums cannot work as all have all muts.
## expect_true(t.test(rowSums(nc1$popSample),rowSums(nc2$popSample),
## alternative = "greater")$p.value < p.value.threshold)
## expect_true(t.test(rowSums(nc2$popSample),rowSums(nc3$popSample),
## alternative = "greater")$p.value < p.value.threshold)
nc1$popSummary[, c(1:3, 8:9)]
nc2$popSummary[, c(1:3, 8:9)]
nc3$popSummary[, c(1:3, 8:9)]
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Similar to above, but mutator has a single element, not the whole
## vector.
cat("\n u7-ossl: a runif is", runif(1), "\n")
pops <- 2000
ft <- 1e-7
lni <- 80
no <- 5e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
pg1 <- runif(lni + 3, min = 1e-7, max = 1e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(pg1) <- sample(names(ni))
mutator1 <- c("oreoisasabgene" = 50) ## a single entry
m1 <- allMutatorEffects(noIntGenes = mutator1)
pg1["hereisoneagene"] <- 1e-3 ## to compare with a laarge one
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
m1.pg1.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
pnom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## ## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## increase mutator, decrease max mu
## similar to oss11 in per-gene-mutation-rates but with the mutator
cat("\n u8-ossl: a runif is", runif(1), "\n")
pops <- 5000
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
m1.pg1.b$popSummary[, c(1:3, 8:9)]
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg1, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg1.b))
## pnom("oreoisasabgene", pg1, no, pops)
## snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## ## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL, Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n mcfu6-ossl: a runif is", runif(1), "\n")
pops <- 2000
ft <- 1e-7
lni <- 80
no <- 2e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 1e-7, max = 1e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 47
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## 1e-3
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
model = "McFL",
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
summary(m1.pg1.b$popSummary[, "NumClones"])
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
enom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("MCFL Relative ordering of number of clones with mutator effects", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Stop on size, and do a fast model (stop shortly after)
## But note we have only four positions left, so difficult to detect
cat("\n mcx1-ossl-ossl: a runif is", runif(1), "\n")
pops <- 160
mu <- 1e-6
ft <- 500
fe <- allFitnessEffects(noIntGenes = c("a" = 0.11,
"b" = 0.11,
"c" = 0.11,
"d" = 0.11))
fm6 <- allMutatorEffects(noIntGenes = c("a" = 30,
"b" = 30,
"c" = 30,
"d" = 30))
nc1 <- oncoSimulSample(pops, mu = mu,
fe, muEF = fm6, finalTime = ft,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e5, model = "McFL",
detectionSize = 1.11e5,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm8 <- allMutatorEffects(noIntGenes = c("a" = 2,
"b" = 2,
"c" = 2,
"d" = 2))
nc2 <- oncoSimulSample(pops, mu = mu,
fe, muEF = fm8, finalTime = ft,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e5, model = "McFL",
detectionSize = 1.11e5,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm7 <- allMutatorEffects(noIntGenes = c("a" = 1e-3,
"b" = 1e-3,
"c" = 1e-3,
"d" = 1e-3))
nc3 <- oncoSimulSample(pops, mu = mu,
fe, muEF = fm7, finalTime = ft,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e5, model = "McFL",
detectionSize = 1.11e5,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
expect_true(wilcox.test(nc1$popSummary[, "NumClones"], nc2$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
expect_true(wilcox.test(nc2$popSummary[, "NumClones"], nc3$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
nc1$popSummary[, c(1:3, 8:9)]
nc2$popSummary[, c(1:3, 8:9)]
nc3$popSummary[, c(1:3, 8:9)]
T1 <- (t.test(rowSums(nc1$popSample),rowSums(nc2$popSample),
alternative = "greater")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(nc2$popSample),rowSums(nc3$popSample),
alternative = "greater")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Relative ordering of number of clones with init mutant of mutator effects and s = 0", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Here stop by time, as s = 0
cat("\n x2cd-ossl-ossl: a runif is", runif(1), "\n")
pops <- 80
ni <- rep(0, 50)
names(ni) <- c("a", "b", "c", "d", paste0("n", 1:46))
fe <- allFitnessEffects(noIntGenes = ni)
fm6 <- allMutatorEffects(noIntGenes = c("a" = .05,
"b" = 1,
"c" = 10,
"d" = 50))
nca <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "a",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncb <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "b",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncc <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "c",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncd <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "d",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
## These are the real tests
T1 <- ( wilcox.test(nca$popSummary[, "NumClones"],
ncb$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(ncb$popSummary[, "NumClones"],
ncc$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T3 <- ( wilcox.test(ncc$popSummary[, "NumClones"],
ncd$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(nca$popSample), rowSums(ncb$popSample),
alternative = "less")$p.value < p.value.threshold)
T5 <- (t.test(rowSums(ncb$popSample), rowSums(ncc$popSample),
alternative = "less")$p.value < p.value.threshold)
T6 <- (t.test(rowSums(ncc$popSample), rowSums(ncd$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL, Expect freq genotypes, mutator and var mut rates, ct mut", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n ct_mcfu6-ossl: a runif is", runif(1), "\n")
pops <- 2500
ft <- 1e-7
lni <- 80
no <- 2e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 1e-5, max = 1e-5) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 47
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## 1e-3
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
model = "McFL",
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
summary(m1.pg1.b$popSummary[, "NumClones"])
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
enom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL, Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n mcfu7-ossl: a runif is", runif(1), "\n")
pops <- 2500
ft <- 3e-7
lni <- 80
no <- 2e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
pg1 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(pg1) <- sample(names(ni))
mutator1 <- c("oreoisasabgene" = 20) ## a single entry
m1 <- allMutatorEffects(noIntGenes = mutator1)
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## 1e-3
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
model = "McFL",
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
summary(m1.pg1.b$popSummary[, "NumClones"])
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
enom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n sameu6-ossl: a runif is", runif(1), "\n")
pops <- 1500
ft <- 1e-5 ## small, as we cannot afford to accumulate many mutations
## or else, given that we have a wholePopulation sample, we
## get the wrong result. Not the case with single cell sampling.
lni <- 70 ##80
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 5e-4, max = 5e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 10 ## 34 ## 53
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## have something huge
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
pnom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n u6-ossl: a runif is", runif(1), "\n")
pops <- 1500
ft <- 1e-5 ## small, as we cannot afford to accumulate many mutations
## or else, given that we have a wholePopulation sample, we
## get the wrong result. Not the case with single cell sampling.
lni <- 70 ##80
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 1e-5, max = 5e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 10 ## 34 ## 53
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## have something huge
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
pnom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("MCFL Relative ordering of number of clones with init mutant of mutator effects and s = 0", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Stopping on time: all s = 0.
cat("\n mcx2cd-ossl: a runif is", runif(1), "\n")
pops <- 40
ni <- rep(0, 50)
names(ni) <- c("a", "b", "c", "d", paste0("n", 1:46))
fe <- allFitnessEffects(noIntGenes = ni)
fm6 <- allMutatorEffects(noIntGenes = c("a" = .05,
"b" = 1,
"c" = 10,
"d" = 50))
nca <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "a", detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
ncb <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "b", detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
ncc <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "c", detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
ncd <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "d",detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
## These are the real tests
T1 <- ( wilcox.test(nca$popSummary[, "NumClones"],
ncb$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(ncb$popSummary[, "NumClones"],
ncc$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T3 <- ( wilcox.test(ncc$popSummary[, "NumClones"],
ncd$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(nca$popSample), rowSums(ncb$popSample),
alternative = "less")$p.value < p.value.threshold)
T5 <- (t.test(rowSums(ncb$popSample), rowSums(ncc$popSample),
alternative = "less")$p.value < p.value.threshold)
T6 <- (t.test(rowSums(ncc$popSample), rowSums(ncd$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Relative ordering of number of clones with init mutant of mutator effects", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## here we do not fill all positions, except maybe for ncd
## We stop on finalTime, not popsize
cat("\n x2bc-ossl: a runif is", runif(1), "\n")
pops <- 30
ni <- rep(0.01, 50)
names(ni) <- c("a", "b", "c", "d", paste0("n", 1:46))
fe <- allFitnessEffects(noIntGenes = ni)
fm6 <- allMutatorEffects(noIntGenes = c("a" = .05,
"b" = 1,
"c" = 10,
"d" = 50))
nca <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "a",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncb <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "b",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncc <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "c",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncd <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "d",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
T1 <- ( wilcox.test(nca$popSummary[, "NumClones"],
ncb$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(ncb$popSummary[, "NumClones"],
ncc$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T3 <- ( wilcox.test(ncc$popSummary[, "NumClones"],
ncd$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(nca$popSample), rowSums(ncb$popSample),
alternative = "less")$p.value < p.value.threshold)
T5 <- (t.test(rowSums(ncb$popSample), rowSums(ncc$popSample),
alternative = "less")$p.value < p.value.threshold)
T6 <- (t.test(rowSums(ncc$popSample), rowSums(ncd$popSample),
alternative = "less")$p.value < p.value.threshold)
nca$popSummary[, c(1:3, 8:9)]
ncb$popSummary[, c(1:3, 8:9)]
ncc$popSummary[, c(1:3, 8:9)]
ncd$popSummary[, c(1:3, 8:9)]
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Same mu vector, different mutator; diffs in number muts, larger t", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## reproduction, death, and double and possibly triple mutants. We
## decrease init pop size to make this fast.
cat("\n nm1-ossl: a runif is", runif(1), "\n")
pops <- 20
ft <- 1
lni <- 100
no <- 1e5
fi <- rep(0, lni)
muvector <- rep(5e-6, lni)
## scrambling names
names(fi) <- replicate(lni,
paste(sample(letters, 12), collapse = ""))
names(muvector) <- sample(names(fi))
## choose something for mutator
mutator10 <- mutator100 <- fi[5]
mutator10[] <- 10
mutator100[] <- 100
fe <- allFitnessEffects(noIntGenes = fi)
m10 <- allMutatorEffects(noIntGenes = mutator10)
m100 <- allMutatorEffects(noIntGenes = mutator100)
pop10 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m10,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
pop100 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m100,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
## number of clones
T1 <- (wilcox.test(NClonesOSS(pop10), NClonesOSS(pop100),
alternative = "less")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(pop10$popSample), rowSums(pop100$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL: Same mu vector, different mutator; diffs in number muts, larger t", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## reproduction, death, and double and possibly triple mutants. We
## decrease init pop size to make this fast.
cat("\n nm3-ossl: a runif is", runif(1), "\n")
pops <- 20
ft <- 1
lni <- 100
no <- 1e5
fi <- rep(0, lni)
muvector <- rep(5e-6, lni)
## scrambling names
names(fi) <- replicate(lni,
paste(sample(letters, 12), collapse = ""))
names(muvector) <- sample(names(fi))
## choose something for mutator
mutator10 <- mutator100 <- fi[5]
mutator10[] <- 10
mutator100[] <- 100
fe <- allFitnessEffects(noIntGenes = fi)
m10 <- allMutatorEffects(noIntGenes = mutator10)
m100 <- allMutatorEffects(noIntGenes = mutator100)
pop10 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m10,
model = "McFL",
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
pop100 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m100,
model = "McFL",
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
T1 <- (wilcox.test(NClonesOSS(pop10), NClonesOSS(pop100),
alternative = "less")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(pop10$popSample), rowSums(pop100$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date() ## Beware: this uses a lot of RAM without the gc()
test_that("McFL: Mutator modules differences", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n MCFLmmd1-ossl: a runif is", runif(1), "\n")
reps <- 10
no <- 5e3
ft <- 100
mu <- 1e-5
lni <- 50
m1 <- 1
m2 <- 25
m3 <- 50
ni <- rep(0, lni)
gn <- paste0("a", 1:lni)
names(ni) <- gn
gn <- paste(gn, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn))
mut2 <- allMutatorEffects(epistasis = c("A" = m2),
geneToModule = c("A" = gn))
mut3 <- allMutatorEffects(epistasis = c("A" = m3),
geneToModule = c("A" = gn))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
model = "McFL",
seed = NULL
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
model = "McFL",
seed = NULL
)
gc()
b3 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut3,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
model = "McFL",
seed = NULL
)
gc()
b3$popSummary[, c(1:3, 8:9)]
b2$popSummary[, c(1:3, 8:9)]
b1$popSummary[, c(1:3, 8:9)]
## mean(rowSums(b3$popSample))
## mean(rowSums(b2$popSample))
## mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
p.fail <- 0.05
T1 <- ( t.test( b3$popSummary[, "NumClones"],
b2$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T3 <- ( t.test( rowSums(b3$popSample) ,
rowSums(b2$popSample), alternative = "greater")$p.value < p.fail)
T4 <- ( t.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
###### The following used to be in the non-long file. But that is just an overkill.
date()
test_that("Mutator increases by given factor with per-gene-mut rates: major axis and chi-sq test", {
## Two cases: mutator and no mutator, with variable mutation rates.
## rates such that rates of no mutator = rates of mutator * mutator.
## Why not compare mutlitplication factor keeping mutation rates
## constant? Because specially with mutator and large diffs in mut
## rates, with oncoSimulSample you undersample variation with
## wholePop, etc.
## Setings similar to oss11 in per-gene-mutation-rates but with the mutator
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n AEu8-ossl: a runif is", runif(1), "\n")
pops <- 200
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
## summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
## sort(enom("oreoisasabgene", pg1, no, pops))
## sort(snomSampl("oreoisasabgene", m1.pg1.b))
## Compare with the expected for this scenario
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
pg2 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3)
names(pg2) <- names(pg1)
m1.pg2.b <- oncoSimulSample(pops,
fe,
mu = pg2,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg2.b$popSummary[, c(1:3, 8:9)]
## summary(m1.pg2.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg2.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg2.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
## sort(enom("oreoisasabgene", pg2, no, pops))
## sort(snomSampl("oreoisasabgene", m1.pg2.b))
p.fail <- 1e-3
T3 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg2.b),
p = pnom("oreoisasabgene", pg2, no, pops))$p.value > p.fail)
## Compare the mutator with the no mutator
T4 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
snomSampl("oreoisasabgene", m1.pg2.b))$p.value > p.fail)
y <- sqrt(snomSampl("oreoisasabgene", m1.pg1.b))
x <- sqrt(snomSampl("oreoisasabgene", m1.pg2.b))
mma <- smatr::ma(y ~ x, slope.test = 1, elev.test = 0) ## From smatr package, for major axis
## intercept not different from 0
T5 <- (mma$elevtest[[1]]$p > p.fail)
T6 <- (mma$slopetest[[1]]$p > p.fail)
if( T1 && T3 && T4 && T5 && T6) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Same mu vector, different mutator; diffs in number muts, tiny t", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Here, there is no reproduction or death. Just mutation. And no double
## mutants either.
## We test:
## - mutator increases mutation rates as seen in:
## - number of clones created
## - number of total mutation events
cat("\n nm0-ossl: a runif is", runif(1), "\n")
pops <- 20
ft <- .0001
lni <- 100
no <- 1e7
fi <- rep(0, lni)
muvector <- rep(5e-6, lni)
## scrambling names
names(fi) <- replicate(lni,
paste(sample(letters, 12), collapse = ""))
names(muvector) <- sample(names(fi))
## choose something for mutator
mutator10 <- mutator100 <- fi[5]
mutator10[] <- 10
mutator100[] <- 100
fe <- allFitnessEffects(noIntGenes = fi)
m10 <- allMutatorEffects(noIntGenes = mutator10)
m100 <- allMutatorEffects(noIntGenes = mutator100)
pop10 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m10,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
pop100 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m100,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
## number of total mutations do not make sense with oncoSimulSample,
## since we cannot estimate them. we approximate wit sum of
## mutations. but that is too thick grain. number of clones is much cleaner
T1 <- (wilcox.test(NClonesOSS(pop10), NClonesOSS(pop100),
alternative = "less")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(pop10$popSample), rowSums(pop100$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that(" Init with different mutators", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n z2-ossl: a runif is", runif(1), "\n")
pops <- 40
ft <- .005
lni <- 50
no <- 1e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- mutator2 <- rep(1, lni + 3)
pg1 <- rep(5e-6, lni + 3)
## scramble names of mutator and per-gene too
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["hereisoneagene"] <- 100
mutator1["oreoisasabgene"] <- 1
mutator1["nnhsisthecgene"] <- 0.01
m1 <- allMutatorEffects(noIntGenes = mutator1)
m1.pg1.a <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = "hereisoneagene",
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = "oreoisasabgene",
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
m1.pg1.c <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = "nnhsisthecgene",
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
T1 <- (wilcox.test(NClonesOSS(m1.pg1.b), NClonesOSS(m1.pg1.a),
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(NClonesOSS(m1.pg1.c), NClonesOSS(m1.pg1.b),
alternative = "less")$p.value < p.value.threshold)
T3 <- (t.test(rowSums(m1.pg1.a$popSample),rowSums(m1.pg1.b$popSample),
alternative = "greater")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(m1.pg1.b$popSample),rowSums(m1.pg1.c$popSample),
alternative = "greater")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
cat(paste("\n Finished test.mutator-oncoSimulSample-long.R test at", date(), "\n"))
Try the OncoSimulR package in your browser
Any scripts or data that you put into this service are public.
OncoSimulR documentation built on Nov. 8, 2020, 8:31 p.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
## Repeat tests in test.mutator, using oncoSimulSample.
## This is a concession to extreme paranoia.
cat(paste("\n Starting test.mutator-oncoSimulSample-long.R test at", date()))
cat(paste("\n a runif ", runif(1), "\n"))
## RNGkind("L'Ecuyer-CMRG") ## for the mclapplies
## require(car) ## for linearHypothesis, below. In the namespace
enom <- function(name, mu, ni = no, pp = pops) {
## expected without a given name for init
ii <- which(names(mu) == name)
out <- ni * pp * mu[-ii]
out[order(names(out))]
}
pnom <- function(name, mu, ni = no, pp = pops) {
ee <- enom(name, mu, ni, pp)
ee/sum(ee)
}
snomSampl <- function(name, out) {
## observed without the init
cs <- colSums(out$popSample)
ii <- which(names(cs) == name)
cs <- cs[-ii]
cs[order(names(cs))]
}
smSampl <- function(name, out) {
## totals for a given gene
cs <- colSums(out$popSample)
ii <- which(names(cs) == name)
cs[ii]
}
totalindSampl <- function(out) {
## total num indivs
sum(out$popSummary$TotalPopSize)
}
medianNClonesOSS <- function(x) {
median(x$popSummary[, "NumClones"])
}
NClonesOSS <- function(x) {
x$popSummary[, "NumClones"]
}
## ugly hack. Of course, not really mutations per clone. But the closest iwth oncoSimulSample and sampling whole pop.
mutsPerCloneOSS <- function(out) {
rowSums(out$popSample)
}
p.value.threshold <- 0.005
## very slow, because huge number of clones. But tests several phenomena comprehensively.
## same with McFL below
date()
test_that("per-gene-mut rates and mutator", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n oss11-ossl: a runif is", runif(1), "\n")
ng <- 40
ni <- rep(0, ng)
m1 <- runif(ng, min = 1e-7, max = 5e-6)
m2 <- rep(1e-5, ng) ## runif(ng, min = 1e-5, max = 1e-4): ## crazy num of clones
names(ni) <- names(m1) <- names(m2) <- c(replicate(ng,
paste(sample(letters, 12), collapse = "")))
fe1 <- allFitnessEffects(noIntGenes = ni)
ft <- 25 ## 50 this is crazy and takes forever
no <- 5e5
reps <- 40
gn <- paste(names(ni), collapse = ", ")
mutator1 <- allMutatorEffects(epistasis = c("MU" = 20),
geneToModule = c("MU" = gn))
mutator2 <- allMutatorEffects(epistasis = c("MU" = 40),
geneToModule = c("MU" = gn))
m1.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m1.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m1.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
runif(1)
m2.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m2.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
m2.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000
)
if(! (
inherits(m1.mutator0$popSummary, "data.frame") &&
inherits(m1.mutator1$popSummary, "data.frame") &&
inherits(m1.mutator2$popSummary, "data.frame") &&
inherits(m2.mutator0$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") ) ) {
T8 <- FALSE
cat("\n not a data frame?\n")
}
if(T8) {
m1.mutator0$popSummary[, c(1:3, 8:9)]
m1.mutator1$popSummary[, c(1:3, 8:9)]
m1.mutator2$popSummary[, c(1:3, 8:9)]
m2.mutator0$popSummary[, c(1:3, 8:9)]
m2.mutator1$popSummary[, c(1:3, 8:9)]
m2.mutator2$popSummary[, c(1:3, 8:9)]
## Mutator increases if larger mutator and compared to no mutator
## within levels of per-gene mutation rates
T1 <- ( wilcox.test(m1.mutator2$popSummary[, "NumClones"], m1.mutator1$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T2 <- ( wilcox.test(m1.mutator1$popSummary[, "NumClones"], m1.mutator0$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T3 <- ( wilcox.test(m2.mutator2$popSummary[, "NumClones"], m2.mutator1$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T4 <- ( wilcox.test(m2.mutator1$popSummary[, "NumClones"], m2.mutator0$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
## Increases in mutation rates increase clones, etc, within levels of
## mutator.
T5 <- ( wilcox.test(m2.mutator0$popSummary[, "NumClones"], m1.mutator0$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T6 <- ( wilcox.test(m2.mutator1$popSummary[, "NumClones"], m1.mutator1$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T7 <- ( wilcox.test(m2.mutator2$popSummary[, "NumClones"], m1.mutator2$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
## expect_true( mean(mutsPerCloneOSS(m1.mutator0)) <
## mean(mutsPerCloneOSS(m2.mutator0)))
## expect_true( mean(mutsPerCloneOSS(m1.mutator1)) <
## mean(mutsPerCloneOSS(m2.mutator1)))
## expect_true( mean(mutsPerCloneOSS(m1.mutator2)) <
## mean(mutsPerCloneOSS(m2.mutator2)))
}
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL: per-gene-mut rates and mutator", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mcfloss11-ossl: a runif is", runif(1), "\n")
ng <- 40
ni <- rep(0, ng)
m1 <- rep(5e-6, ng) ## too much variation and hard to pick the diffs.;
## runif(ng, min = 1e-7, max = 5e-6) And if too
## tiny, you do not pick them up unless huge ft
## and then it is way too slow for m2, etc.
m2 <- rep(5e-5, ng)
names(ni) <- names(m1) <- names(m2) <- c(replicate(ng,
paste(sample(letters, 12), collapse = "")))
fe1 <- allFitnessEffects(noIntGenes = ni)
ft <- 20 ## 50
no <- 5e5
reps <- 20 ## 40
gn <- paste(names(ni), collapse = ", ")
mutator1 <- allMutatorEffects(epistasis = c("MU" = 20),
geneToModule = c("MU" = gn))
mutator2 <- allMutatorEffects(epistasis = c("MU" = 40),
geneToModule = c("MU" = gn))
m1.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m1.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m1.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m1,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
cat("\n starting m2\n")
m2.mutator0 <- oncoSimulSample(reps,
fe1,
mu = m2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m2.mutator1 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
m2.mutator2 <- oncoSimulSample(reps,
fe1,
mu = m2,
muEF = mutator2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, max.wall.time = 2000, model = "McFL"
)
if(! (
inherits(m1.mutator0$popSummary, "data.frame") &&
inherits(m1.mutator1$popSummary, "data.frame") &&
inherits(m1.mutator2$popSummary, "data.frame") &&
inherits(m2.mutator0$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") &&
inherits(m2.mutator1$popSummary, "data.frame") ) ) {
cat("\n not a data frame?\n")
T8 <- FALSE
}
if(T8) {
m1.mutator0$popSummary[, c(1:3, 8:9)]
m1.mutator1$popSummary[, c(1:3, 8:9)]
m1.mutator2$popSummary[, c(1:3, 8:9)]
m2.mutator0$popSummary[, c(1:3, 8:9)]
m2.mutator1$popSummary[, c(1:3, 8:9)]
m2.mutator2$popSummary[, c(1:3, 8:9)]
## Mutator increases if larger mutator and compared to no mutator
## within levels of per-gene mutation rates
## we could use wilcoxon or t tests actually, specially because often diffs
## are not huge.
p.fail <- 0.005
T1 <- ( t.test( m1.mutator2$popSummary[, "NumClones"] ,
m1.mutator1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( m1.mutator1$popSummary[, "NumClones"] ,
m1.mutator0$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T3 <- ( t.test( m2.mutator2$popSummary[, "NumClones"] ,
m2.mutator1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T4 <- ( t.test( m2.mutator1$popSummary[, "NumClones"] ,
m2.mutator0$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
## Increases in mutation rates increase clones, etc, within levels of
## mutator.
T5 <- ( t.test( m1.mutator0$popSummary[, "NumClones"] ,
m2.mutator0$popSummary[, "NumClones"], alternative = "less")$p.value < p.fail)
T6 <- ( t.test( m1.mutator1$popSummary[, "NumClones"] ,
m2.mutator1$popSummary[, "NumClones"], alternative = "less")$p.value < p.fail)
T7 <- ( t.test( m1.mutator2$popSummary[, "NumClones"] ,
m2.mutator2$popSummary[, "NumClones"], alternative = "less")$p.value < p.fail)
}
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Mutator increases by given factor with per-gene-mut rates: major axis and chi-sq test", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Two cases: mutator and no mutator, with variable mutation rates.
## rates such that rates of no mutator = rates of mutator * mutator.
## Why not compare mutlitplication factor keeping mutation rates
## constant? Because specially with mutator and large diffs in mut
## rates, with oncoSimulSample you undersample variation with
## wholePop, etc.
## Setings similar to oss11 in per-gene-mutation-rates but with the mutator
cat("\n AEu8_long-ossl: a runif is", runif(1), "\n")
pops <- 8000
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg1, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg1.b))
## Compare with the expected for this scenario
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
pg2 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3)
names(pg2) <- names(pg1)
m1.pg2.b <- oncoSimulSample(pops,
fe,
mu = pg2,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg2.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg2.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg2.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg2.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg2, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg2.b))
p.fail <- 1e-3
T2 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg2.b),
p = pnom("oreoisasabgene", pg2, no, pops))$p.value > p.fail)
## Compare the mutator with the no mutator
T3 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
snomSampl("oreoisasabgene", m1.pg2.b))$p.value > p.fail)
y <- sqrt(snomSampl("oreoisasabgene", m1.pg1.b))
x <- sqrt(snomSampl("oreoisasabgene", m1.pg2.b))
mma <- smatr::ma(y ~ x, slope.test = 1, elev.test = 0) ## From smatr package, for major axis
## intercept not different from 0
T4 <- (mma$elevtest[[1]]$p > p.fail)
T5 <- (mma$slopetest[[1]]$p > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL: Mutator increases by given factor with per-gene-mut rates: major axis and chi-sq test", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Two cases: mutator and no mutator, with variable mutation rates.
## rates such that rates of no mutator = rates of mutator * mutator.
## Why not compare mutlitplication factor keeping mutation rates
## constant? Because specially with mutator and large diffs in mut
## rates, with oncoSimulSample you undersample variation with
## wholePop, etc.
## Setings similar to oss11 in per-gene-mutation-rates but with the mutator
cat("\n MCFL: long_AEu8-ossl: a runif is", runif(1), "\n")
pops <- 8000
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
model = "McFL",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg1, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg1.b))
## Compare with the expected for this scenario
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
pg2 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3)
names(pg2) <- names(pg1)
m1.pg2.b <- oncoSimulSample(pops,
fe,
mu = pg2,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
model = "McFL",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg2.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg2.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg2.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg2.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg2, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg2.b))
p.fail <- 1e-3
T2 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg2.b),
p = pnom("oreoisasabgene", pg2, no, pops))$p.value > p.fail)
## Compare mutator with no mutator
T3 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
snomSampl("oreoisasabgene", m1.pg2.b))$p.value > p.fail)
y <- sqrt(snomSampl("oreoisasabgene", m1.pg1.b))
x <- sqrt(snomSampl("oreoisasabgene", m1.pg2.b))
mma <- smatr::ma(y ~ x, slope.test = 1, elev.test = 0) ## From smatr package, for major axis
## intercept not different from 0
T4 <- (mma$elevtest[[1]]$p > p.fail)
T5 <- (mma$slopetest[[1]]$p > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Mutator, several modules differences", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mmd1_2-ossl: a runif is", runif(1), "\n")
reps <- 140
no <- 5e3
ft <- 80 ## you need it large enough to get enough hits
mu <- 1e-5
ln <- 50
m1 <- 7 ## if this is too large, easy to get it to blow.
ni <- rep(0, 2 * ln)
gna <- paste0("a", 1:ln)
gnb <- paste0("b", 1:ln)
names(ni) <- c(gna, gnb)
gn1 <- paste(c(gna, gnb), collapse = ", ")
gna <- paste(gna, collapse = ", ")
gnb <- paste(gnb, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn1))
mut2 <- allMutatorEffects(epistasis = c("A" = m1,
"B" = m1),
geneToModule = c("A" = gna,
"B" = gnb))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
## b2$popSummary[, c(1:3, 8:9)]
## b1$popSummary[, c(1:3, 8:9)]
## mean(rowSums(b2$popSample))
## mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
## Variances for NumClones are hugely unequal, even after log transform.;
## might want Wilcoxon? Similar for rowSums of popSample
p.fail <- 0.05
T1 <- ( wilcox.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( wilcox.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
## Remember that numClones is underestimated, possibly severly, by
## oncoSimulSample compared to oncoSimulPop, since we only look at the
## clones that exist at the end.
date()
test_that("Mutator, several modules differences, McFL", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mmd1_2_mc-ossl: a runif is", runif(1), "\n")
reps <- 80
no <- 5e3
ft <- 50 ## you need it large enough to get enough hits
mu <- 1e-5
ln <- 50
m1 <- 7 ## if this is too large, easy to get it to blow.
ni <- rep(0, 2 * ln)
gna <- paste0("a", 1:ln)
gnb <- paste0("b", 1:ln)
names(ni) <- c(gna, gnb)
gn1 <- paste(c(gna, gnb), collapse = ", ")
gna <- paste(gna, collapse = ", ")
gnb <- paste(gnb, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn1))
mut2 <- allMutatorEffects(epistasis = c("A" = m1,
"B" = m1),
geneToModule = c("A" = gna,
"B" = gnb))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, model = "McFL"
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, model = "McFL"
)
gc()
b2$popSummary[, c(1:3, 8:9)]
b1$popSummary[, c(1:3, 8:9)]
mean(rowSums(b2$popSample))
mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
p.fail <- 0.05
T1 <- ( t.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
## Slow (~ 3 seconds) but tests modules of mutator nicely.
date() ## Beware: this uses a lot of RAM without the gc()
test_that("Mutator modules differences", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n mmd1-ossl: a runif is", runif(1), "\n")
reps <- 40
no <- 5e3
ft <- 100
mu <- 1e-5
lni <- 50
m1 <- 1
m2 <- 25
m3 <- 50
ni <- rep(0, lni)
gn <- paste0("a", 1:lni)
names(ni) <- gn
gn <- paste(gn, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn))
mut2 <- allMutatorEffects(epistasis = c("A" = m2),
geneToModule = c("A" = gn))
mut3 <- allMutatorEffects(epistasis = c("A" = m3),
geneToModule = c("A" = gn))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b3 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut3,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL
)
gc()
b3$popSummary[, c(1:3, 8:9)]
b2$popSummary[, c(1:3, 8:9)]
b1$popSummary[, c(1:3, 8:9)]
## mean(rowSums(b3$popSample))
## mean(rowSums(b2$popSample))
## mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
p.fail <- 0.05
T1 <- ( t.test( b3$popSummary[, "NumClones"],
b2$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T3 <- ( t.test( rowSums(b3$popSample) ,
rowSums(b2$popSample), alternative = "greater")$p.value < p.fail)
T4 <- ( t.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Relative ordering of number of clones with mutator effects", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n x1-ossl: a runif is", runif(1), "\n")
pops <- 40
fe <- allFitnessEffects(noIntGenes = c("a" = 0.12,
"b" = 0.14,
"c" = 0.16,
"d" = 0.11))
fm6 <- allMutatorEffects(noIntGenes = c("a" = 5,
"b" = 5,
"c" = 5,
"d" = 5))
nc1 <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =250,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e6,
detectionSize = 1e8,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm8 <- allMutatorEffects(noIntGenes = c("a" = 1,
"b" = 1,
"c" = 1,
"d" = 1))
nc2 <- oncoSimulSample(pops, fe, muEF = fm8, finalTime =250,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e6,
detectionSize = 1e8,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm7 <- allMutatorEffects(noIntGenes = c("a" = 1e-3,
"b" = 1e-3,
"c" = 1e-3,
"d" = 1e-3))
nc3 <- oncoSimulSample(pops, fe, muEF = fm7, finalTime =250,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e6,
detectionSize = 1e8,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
T1 <- (wilcox.test(nc1$popSummary[, "NumClones"], nc2$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
T2 <- (wilcox.test(nc2$popSummary[, "NumClones"], nc3$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
## rowSums cannot work as all have all muts.
## expect_true(t.test(rowSums(nc1$popSample),rowSums(nc2$popSample),
## alternative = "greater")$p.value < p.value.threshold)
## expect_true(t.test(rowSums(nc2$popSample),rowSums(nc3$popSample),
## alternative = "greater")$p.value < p.value.threshold)
nc1$popSummary[, c(1:3, 8:9)]
nc2$popSummary[, c(1:3, 8:9)]
nc3$popSummary[, c(1:3, 8:9)]
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Similar to above, but mutator has a single element, not the whole
## vector.
cat("\n u7-ossl: a runif is", runif(1), "\n")
pops <- 2000
ft <- 1e-7
lni <- 80
no <- 5e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
pg1 <- runif(lni + 3, min = 1e-7, max = 1e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(pg1) <- sample(names(ni))
mutator1 <- c("oreoisasabgene" = 50) ## a single entry
m1 <- allMutatorEffects(noIntGenes = mutator1)
pg1["hereisoneagene"] <- 1e-3 ## to compare with a laarge one
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
m1.pg1.b$popSummary[, c(1:3, 8:9)]
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
pnom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## ## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## increase mutator, decrease max mu
## similar to oss11 in per-gene-mutation-rates but with the mutator
cat("\n u8-ossl: a runif is", runif(1), "\n")
pops <- 5000
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
m1.pg1.b$popSummary[, c(1:3, 8:9)]
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
sort(enom("oreoisasabgene", pg1, no, pops))
sort(snomSampl("oreoisasabgene", m1.pg1.b))
## pnom("oreoisasabgene", pg1, no, pops)
## snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## ## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL, Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n mcfu6-ossl: a runif is", runif(1), "\n")
pops <- 2000
ft <- 1e-7
lni <- 80
no <- 2e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 1e-7, max = 1e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 47
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## 1e-3
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
model = "McFL",
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
summary(m1.pg1.b$popSummary[, "NumClones"])
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
enom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("MCFL Relative ordering of number of clones with mutator effects", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Stop on size, and do a fast model (stop shortly after)
## But note we have only four positions left, so difficult to detect
cat("\n mcx1-ossl-ossl: a runif is", runif(1), "\n")
pops <- 160
mu <- 1e-6
ft <- 500
fe <- allFitnessEffects(noIntGenes = c("a" = 0.11,
"b" = 0.11,
"c" = 0.11,
"d" = 0.11))
fm6 <- allMutatorEffects(noIntGenes = c("a" = 30,
"b" = 30,
"c" = 30,
"d" = 30))
nc1 <- oncoSimulSample(pops, mu = mu,
fe, muEF = fm6, finalTime = ft,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e5, model = "McFL",
detectionSize = 1.11e5,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm8 <- allMutatorEffects(noIntGenes = c("a" = 2,
"b" = 2,
"c" = 2,
"d" = 2))
nc2 <- oncoSimulSample(pops, mu = mu,
fe, muEF = fm8, finalTime = ft,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e5, model = "McFL",
detectionSize = 1.11e5,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
fm7 <- allMutatorEffects(noIntGenes = c("a" = 1e-3,
"b" = 1e-3,
"c" = 1e-3,
"d" = 1e-3))
nc3 <- oncoSimulSample(pops, mu = mu,
fe, muEF = fm7, finalTime = ft,
mutationPropGrowth = FALSE,
sampleEvery = 0.01, thresholdWhole = 1e-20,
initSize = 1e5, model = "McFL",
detectionSize = 1.11e5,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
expect_true(wilcox.test(nc1$popSummary[, "NumClones"], nc2$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
expect_true(wilcox.test(nc2$popSummary[, "NumClones"], nc3$popSummary[, "NumClones"],
alternative = "greater")$p.value < p.value.threshold)
nc1$popSummary[, c(1:3, 8:9)]
nc2$popSummary[, c(1:3, 8:9)]
nc3$popSummary[, c(1:3, 8:9)]
T1 <- (t.test(rowSums(nc1$popSample),rowSums(nc2$popSample),
alternative = "greater")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(nc2$popSample),rowSums(nc3$popSample),
alternative = "greater")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Relative ordering of number of clones with init mutant of mutator effects and s = 0", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Here stop by time, as s = 0
cat("\n x2cd-ossl-ossl: a runif is", runif(1), "\n")
pops <- 80
ni <- rep(0, 50)
names(ni) <- c("a", "b", "c", "d", paste0("n", 1:46))
fe <- allFitnessEffects(noIntGenes = ni)
fm6 <- allMutatorEffects(noIntGenes = c("a" = .05,
"b" = 1,
"c" = 10,
"d" = 50))
nca <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "a",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncb <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "b",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncc <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "c",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncd <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "d",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
## These are the real tests
T1 <- ( wilcox.test(nca$popSummary[, "NumClones"],
ncb$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(ncb$popSummary[, "NumClones"],
ncc$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T3 <- ( wilcox.test(ncc$popSummary[, "NumClones"],
ncd$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(nca$popSample), rowSums(ncb$popSample),
alternative = "less")$p.value < p.value.threshold)
T5 <- (t.test(rowSums(ncb$popSample), rowSums(ncc$popSample),
alternative = "less")$p.value < p.value.threshold)
T6 <- (t.test(rowSums(ncc$popSample), rowSums(ncd$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL, Expect freq genotypes, mutator and var mut rates, ct mut", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n ct_mcfu6-ossl: a runif is", runif(1), "\n")
pops <- 2500
ft <- 1e-7
lni <- 80
no <- 2e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 1e-5, max = 1e-5) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 47
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## 1e-3
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
model = "McFL",
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
summary(m1.pg1.b$popSummary[, "NumClones"])
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
enom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL, Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n mcfu7-ossl: a runif is", runif(1), "\n")
pops <- 2500
ft <- 3e-7
lni <- 80
no <- 2e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
pg1 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(pg1) <- sample(names(ni))
mutator1 <- c("oreoisasabgene" = 20) ## a single entry
m1 <- allMutatorEffects(noIntGenes = mutator1)
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## 1e-3
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
model = "McFL",
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
summary(m1.pg1.b$popSummary[, "NumClones"])
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
enom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n sameu6-ossl: a runif is", runif(1), "\n")
pops <- 1500
ft <- 1e-5 ## small, as we cannot afford to accumulate many mutations
## or else, given that we have a wholePopulation sample, we
## get the wrong result. Not the case with single cell sampling.
lni <- 70 ##80
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 5e-4, max = 5e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 10 ## 34 ## 53
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## have something huge
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
pnom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Expect freq genotypes, mutator and var mut rates", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## We test that mutator does not affect expected frequencies of
## mutated genes: they are given by the mutation rate of each gene.
cat("\n u6-ossl: a runif is", runif(1), "\n")
pops <- 1500
ft <- 1e-5 ## small, as we cannot afford to accumulate many mutations
## or else, given that we have a wholePopulation sample, we
## get the wrong result. Not the case with single cell sampling.
lni <- 70 ##80
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
## of course, passing a mutator of 1 makes everything slow.
mutator1 <- rep(1, lni + 3)
## pg1 <- rep(1e-5, lni + 3)
pg1 <- runif(lni + 3, min = 1e-5, max = 5e-4) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 10 ## 34 ## 53
m1 <- allMutatorEffects(noIntGenes = mutator1)
## have something with much larger mutation rate
pg1["hereisoneagene"] <- 1e-3 ## have something huge
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## If numclones is much larger than 2, that signals trouble as you are
## smoothing differences between frequencies with oncoSimulSample,
## whole pop
summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
pnom("oreoisasabgene", pg1, no, pops)
snomSampl("oreoisasabgene", m1.pg1.b)
## plot(snomSampl("oreoisasabgene", m1.pg1.b)/sum(snomSampl("oreoisasabgene", m1.pg1.b)) ~
## pnom("oreoisasabgene", pg1, no, pops)); abline(a = 0, b = 1)
## yes, if very large prob for one, it is slightly underestimated
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("MCFL Relative ordering of number of clones with init mutant of mutator effects and s = 0", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Stopping on time: all s = 0.
cat("\n mcx2cd-ossl: a runif is", runif(1), "\n")
pops <- 40
ni <- rep(0, 50)
names(ni) <- c("a", "b", "c", "d", paste0("n", 1:46))
fe <- allFitnessEffects(noIntGenes = ni)
fm6 <- allMutatorEffects(noIntGenes = c("a" = .05,
"b" = 1,
"c" = 10,
"d" = 50))
nca <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "a", detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
ncb <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "b", detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
ncc <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "c", detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
ncd <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "d",detectionSize = 1e9,
sampleEvery = 0.01, thresholdWhole = 1e-20,
onlyCancer = FALSE, detectionProb = NA, model = "McFL")
## These are the real tests
T1 <- ( wilcox.test(nca$popSummary[, "NumClones"],
ncb$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(ncb$popSummary[, "NumClones"],
ncc$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T3 <- ( wilcox.test(ncc$popSummary[, "NumClones"],
ncd$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(nca$popSample), rowSums(ncb$popSample),
alternative = "less")$p.value < p.value.threshold)
T5 <- (t.test(rowSums(ncb$popSample), rowSums(ncc$popSample),
alternative = "less")$p.value < p.value.threshold)
T6 <- (t.test(rowSums(ncc$popSample), rowSums(ncd$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Relative ordering of number of clones with init mutant of mutator effects", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## here we do not fill all positions, except maybe for ncd
## We stop on finalTime, not popsize
cat("\n x2bc-ossl: a runif is", runif(1), "\n")
pops <- 30
ni <- rep(0.01, 50)
names(ni) <- c("a", "b", "c", "d", paste0("n", 1:46))
fe <- allFitnessEffects(noIntGenes = ni)
fm6 <- allMutatorEffects(noIntGenes = c("a" = .05,
"b" = 1,
"c" = 10,
"d" = 50))
nca <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "a",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncb <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "b",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncc <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "c",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
ncd <- oncoSimulSample(pops, fe, muEF = fm6, finalTime =50,
mutationPropGrowth = FALSE,
initSize = 1e4,
initMutant = "d",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999, seed = NULL,
onlyCancer = FALSE, detectionProb = NA)
T1 <- ( wilcox.test(nca$popSummary[, "NumClones"],
ncb$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(ncb$popSummary[, "NumClones"],
ncc$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T3 <- ( wilcox.test(ncc$popSummary[, "NumClones"],
ncd$popSummary[, "NumClones"],
alternative = "less")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(nca$popSample), rowSums(ncb$popSample),
alternative = "less")$p.value < p.value.threshold)
T5 <- (t.test(rowSums(ncb$popSample), rowSums(ncc$popSample),
alternative = "less")$p.value < p.value.threshold)
T6 <- (t.test(rowSums(ncc$popSample), rowSums(ncd$popSample),
alternative = "less")$p.value < p.value.threshold)
nca$popSummary[, c(1:3, 8:9)]
ncb$popSummary[, c(1:3, 8:9)]
ncc$popSummary[, c(1:3, 8:9)]
ncd$popSummary[, c(1:3, 8:9)]
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Same mu vector, different mutator; diffs in number muts, larger t", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## reproduction, death, and double and possibly triple mutants. We
## decrease init pop size to make this fast.
cat("\n nm1-ossl: a runif is", runif(1), "\n")
pops <- 20
ft <- 1
lni <- 100
no <- 1e5
fi <- rep(0, lni)
muvector <- rep(5e-6, lni)
## scrambling names
names(fi) <- replicate(lni,
paste(sample(letters, 12), collapse = ""))
names(muvector) <- sample(names(fi))
## choose something for mutator
mutator10 <- mutator100 <- fi[5]
mutator10[] <- 10
mutator100[] <- 100
fe <- allFitnessEffects(noIntGenes = fi)
m10 <- allMutatorEffects(noIntGenes = mutator10)
m100 <- allMutatorEffects(noIntGenes = mutator100)
pop10 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m10,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
pop100 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m100,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
## number of clones
T1 <- (wilcox.test(NClonesOSS(pop10), NClonesOSS(pop100),
alternative = "less")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(pop10$popSample), rowSums(pop100$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("McFL: Same mu vector, different mutator; diffs in number muts, larger t", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## reproduction, death, and double and possibly triple mutants. We
## decrease init pop size to make this fast.
cat("\n nm3-ossl: a runif is", runif(1), "\n")
pops <- 20
ft <- 1
lni <- 100
no <- 1e5
fi <- rep(0, lni)
muvector <- rep(5e-6, lni)
## scrambling names
names(fi) <- replicate(lni,
paste(sample(letters, 12), collapse = ""))
names(muvector) <- sample(names(fi))
## choose something for mutator
mutator10 <- mutator100 <- fi[5]
mutator10[] <- 10
mutator100[] <- 100
fe <- allFitnessEffects(noIntGenes = fi)
m10 <- allMutatorEffects(noIntGenes = mutator10)
m100 <- allMutatorEffects(noIntGenes = mutator100)
pop10 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m10,
model = "McFL",
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
pop100 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m100,
model = "McFL",
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
T1 <- (wilcox.test(NClonesOSS(pop10), NClonesOSS(pop100),
alternative = "less")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(pop10$popSample), rowSums(pop100$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date() ## Beware: this uses a lot of RAM without the gc()
test_that("McFL: Mutator modules differences", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n MCFLmmd1-ossl: a runif is", runif(1), "\n")
reps <- 10
no <- 5e3
ft <- 100
mu <- 1e-5
lni <- 50
m1 <- 1
m2 <- 25
m3 <- 50
ni <- rep(0, lni)
gn <- paste0("a", 1:lni)
names(ni) <- gn
gn <- paste(gn, collapse = ", ")
mut1 <- allMutatorEffects(epistasis = c("A" = m1),
geneToModule = c("A" = gn))
mut2 <- allMutatorEffects(epistasis = c("A" = m2),
geneToModule = c("A" = gn))
mut3 <- allMutatorEffects(epistasis = c("A" = m3),
geneToModule = c("A" = gn))
f1 <- allFitnessEffects(noIntGenes = ni)
b1 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut1,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
model = "McFL",
seed = NULL
)
gc()
b2 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut2,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
model = "McFL",
seed = NULL
)
gc()
b3 <- oncoSimulSample(reps,
f1,
mu = mu,
muEF = mut3,
onlyCancer = FALSE, detectionProb = NA,
initSize = no,
finalTime = ft,
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
model = "McFL",
seed = NULL
)
gc()
b3$popSummary[, c(1:3, 8:9)]
b2$popSummary[, c(1:3, 8:9)]
b1$popSummary[, c(1:3, 8:9)]
## mean(rowSums(b3$popSample))
## mean(rowSums(b2$popSample))
## mean(rowSums(b1$popSample))
## This is, of course, affected by sampling only at end: we do not see
## the many intermediate events.
p.fail <- 0.05
T1 <- ( t.test( b3$popSummary[, "NumClones"],
b2$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T2 <- ( t.test( b2$popSummary[, "NumClones"],
b1$popSummary[, "NumClones"], alternative = "greater")$p.value < p.fail)
T3 <- ( t.test( rowSums(b3$popSample) ,
rowSums(b2$popSample), alternative = "greater")$p.value < p.fail)
T4 <- ( t.test( rowSums(b2$popSample) ,
rowSums(b1$popSample), alternative = "greater")$p.value < p.fail)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
###### The following used to be in the non-long file. But that is just an overkill.
date()
test_that("Mutator increases by given factor with per-gene-mut rates: major axis and chi-sq test", {
## Two cases: mutator and no mutator, with variable mutation rates.
## rates such that rates of no mutator = rates of mutator * mutator.
## Why not compare mutlitplication factor keeping mutation rates
## constant? Because specially with mutator and large diffs in mut
## rates, with oncoSimulSample you undersample variation with
## wholePop, etc.
## Setings similar to oss11 in per-gene-mutation-rates but with the mutator
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n AEu8-ossl: a runif is", runif(1), "\n")
pops <- 200
ft <- 5e-3
lni <- 7
no <- 5e5
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- rep(1, lni + 3)
pg1 <- seq(from = 1e-9, to = 1e-6, length.out = lni + 3) ## max should not be
## huge here as mutator
## is 34. Can get beyond
## 1
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["oreoisasabgene"] <- 100
m1 <- allMutatorEffects(noIntGenes = mutator1)
## pg1["hereisoneagene"] <- 1e-4 ## if this gets huge, then you are
## ## undersampling and the chi-square will
## ## fail. But then, we probably are
## ## running into numerical issues: 3
## ## orders of magnitude differences.
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg1.b$popSummary[, c(1:3, 8:9)]
## summary(m1.pg1.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg1.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg1.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
## sort(enom("oreoisasabgene", pg1, no, pops))
## sort(snomSampl("oreoisasabgene", m1.pg1.b))
## Compare with the expected for this scenario
p.fail <- 1e-3
T1 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
p = pnom("oreoisasabgene", pg1, no, pops))$p.value > p.fail)
pg2 <- seq(from = 1e-7, to = 1e-4, length.out = lni + 3)
names(pg2) <- names(pg1)
m1.pg2.b <- oncoSimulSample(pops,
fe,
mu = pg2,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant ="oreoisasabgene",
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
onlyCancer = FALSE, detectionProb = NA, seed = NULL)
## m1.pg2.b$popSummary[, c(1:3, 8:9)]
## summary(m1.pg2.b$popSummary[, "NumClones"])
## Recall that init-mutant tests check always present of initMutant
## against a thresholWhole of 1. Here it is slightly different.
expect_true(smSampl("oreoisasabgene", m1.pg2.b) == pops)
## catch a pattern that would make the previous trivially true
expect_false(sum(m1.pg2.b$popSample) == pops * (lni + 3))
## next two, to compare with oss1a
## sort(enom("oreoisasabgene", pg2, no, pops))
## sort(snomSampl("oreoisasabgene", m1.pg2.b))
p.fail <- 1e-3
T3 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg2.b),
p = pnom("oreoisasabgene", pg2, no, pops))$p.value > p.fail)
## Compare the mutator with the no mutator
T4 <- (chisq.test(snomSampl("oreoisasabgene", m1.pg1.b),
snomSampl("oreoisasabgene", m1.pg2.b))$p.value > p.fail)
y <- sqrt(snomSampl("oreoisasabgene", m1.pg1.b))
x <- sqrt(snomSampl("oreoisasabgene", m1.pg2.b))
mma <- smatr::ma(y ~ x, slope.test = 1, elev.test = 0) ## From smatr package, for major axis
## intercept not different from 0
T5 <- (mma$elevtest[[1]]$p > p.fail)
T6 <- (mma$slopetest[[1]]$p > p.fail)
if( T1 && T3 && T4 && T5 && T6) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that("Same mu vector, different mutator; diffs in number muts, tiny t", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
## Here, there is no reproduction or death. Just mutation. And no double
## mutants either.
## We test:
## - mutator increases mutation rates as seen in:
## - number of clones created
## - number of total mutation events
cat("\n nm0-ossl: a runif is", runif(1), "\n")
pops <- 20
ft <- .0001
lni <- 100
no <- 1e7
fi <- rep(0, lni)
muvector <- rep(5e-6, lni)
## scrambling names
names(fi) <- replicate(lni,
paste(sample(letters, 12), collapse = ""))
names(muvector) <- sample(names(fi))
## choose something for mutator
mutator10 <- mutator100 <- fi[5]
mutator10[] <- 10
mutator100[] <- 100
fe <- allFitnessEffects(noIntGenes = fi)
m10 <- allMutatorEffects(noIntGenes = mutator10)
m100 <- allMutatorEffects(noIntGenes = mutator100)
pop10 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m10,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
pop100 <- oncoSimulSample(pops,
fe,
mu = muvector,
muEF = m100,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = names(mutator10),
sampleEvery = 0.01, thresholdWhole = 1e-20,
detectionSize = 1e9,
detectionDrivers = 9999,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
## number of total mutations do not make sense with oncoSimulSample,
## since we cannot estimate them. we approximate wit sum of
## mutations. but that is too thick grain. number of clones is much cleaner
T1 <- (wilcox.test(NClonesOSS(pop10), NClonesOSS(pop100),
alternative = "less")$p.value < p.value.threshold)
T2 <- (t.test(rowSums(pop10$popSample), rowSums(pop100$popSample),
alternative = "less")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
date()
test_that(" Init with different mutators", {
max.tries <- 4
for(tries in 1:max.tries) {
T1 <- T2 <- T3 <- T4 <- T5 <- T6 <- T7 <- T8 <- TRUE
cat("\n z2-ossl: a runif is", runif(1), "\n")
pops <- 40
ft <- .005
lni <- 50
no <- 1e7
ni <- c(0, 0, 0, rep(0, lni))
## scramble around names
names(ni) <- c("hereisoneagene",
"oreoisasabgene",
"nnhsisthecgene",
replicate(lni,
paste(sample(letters, 12), collapse = "")))
ni <- ni[order(names(ni))]
fe <- allFitnessEffects(noIntGenes = ni)
mutator1 <- mutator2 <- rep(1, lni + 3)
pg1 <- rep(5e-6, lni + 3)
## scramble names of mutator and per-gene too
names(mutator1) <- sample(names(ni))
names(pg1) <- sample(names(ni))
mutator1["hereisoneagene"] <- 100
mutator1["oreoisasabgene"] <- 1
mutator1["nnhsisthecgene"] <- 0.01
m1 <- allMutatorEffects(noIntGenes = mutator1)
m1.pg1.a <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = "hereisoneagene",
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
m1.pg1.b <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = "oreoisasabgene",
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
m1.pg1.c <- oncoSimulSample(pops,
fe,
mu = pg1,
muEF = m1,
finalTime = ft,
mutationPropGrowth = FALSE,
initSize = no,
initMutant = "nnhsisthecgene",
detectionSize = 1e9,
detectionDrivers = 9999,
sampleEvery = 0.01, thresholdWhole = 1e-20,
seed = NULL, onlyCancer = FALSE, detectionProb = NA)
T1 <- (wilcox.test(NClonesOSS(m1.pg1.b), NClonesOSS(m1.pg1.a),
alternative = "less")$p.value < p.value.threshold)
T2 <- (wilcox.test(NClonesOSS(m1.pg1.c), NClonesOSS(m1.pg1.b),
alternative = "less")$p.value < p.value.threshold)
T3 <- (t.test(rowSums(m1.pg1.a$popSample),rowSums(m1.pg1.b$popSample),
alternative = "greater")$p.value < p.value.threshold)
T4 <- (t.test(rowSums(m1.pg1.b$popSample),rowSums(m1.pg1.c$popSample),
alternative = "greater")$p.value < p.value.threshold)
if(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8) break;
}
cat(paste("\n done tries", tries, "\n"))
expect_true(T1 && T2 && T3 && T4 && T5 && T6 && T7 && T8)
})
date()
cat(paste("\n Finished test.mutator-oncoSimulSample-long.R test at", date(), "\n"))
Try the OncoSimulR package in your browser
Any scripts or data that you put into this service are public.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Embedding an R snippet on your website
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.