R/modules_relations.r
In hloefflerwirth/oposSOM: Comprehensive analysis of transcriptome data
Defines functions
modules.relations
modules.relations <- function(env, spot.list, main, path)
{
sd.theshold <- sd(spot.list$spotdata)
spotdata.binary <- spot.list$spotdata > sd.theshold
mode(spotdata.binary) <- "numeric"
pdf(path, 29.7/2.54, 21/2.54, useDingbats=FALSE)
#### wTO network ####
if ( length(spot.list$spots) > 2 )
{
# calculate wTO #
adj_matrix <- cor(t(spot.list$spotdata),t(spot.list$spotdata))
diag(adj_matrix) <- 0
omega <- matrix(NA,
nrow=nrow(spot.list$spotdata),
ncol=nrow(spot.list$spotdata),
dimnames=list(names(spot.list$spots),names(spot.list$spots)))
b <- adj_matrix %*% adj_matrix
k_sum <- apply(adj_matrix, 2, function(x) { sum(abs(x)) })
for (i in 1: nrow(spot.list$spotdata))
{
for (j in 1: nrow(spot.list$spotdata))
{
omega[i,j] <-
(b[i,j] + adj_matrix[i,j]) /
(min(k_sum[i],k_sum[j]) + 1 - abs(adj_matrix[i,j]))
}
}
diag(omega) <- 0
omega[which(abs(omega) < 0.25)] <- 0
g <- graph.adjacency(omega, weighted=TRUE, mode="undirected")
V(g)$label <- names(spot.list$spots)
n.spot <- apply(spot.list$spotdata, 1, function(x) sum(x>sd.theshold) )
if (max(n.spot) == min(n.spot))
{
V(g)$size <- 10
} else
{
V(g)$size <- 8 * (n.spot-min(n.spot)) / (max(n.spot)-min(n.spot)) + 6
}
if (length(E(g)) > 0)
{
E(g)$color <- c("brown4","chocolate","rosybrown1","white","white","lightblue","cornflowerblue","dodgerblue4")[ cut( E(g)$weight, breaks=c(-Inf,-0.75,-0.5,-0.25,0,0.25,0.5,0.75,Inf), labels=c(1:8) ) ]
E(g)$width <- abs(E(g)$weight)
if (max(E(g)$width) == min(E(g)$width))
{
E(g)$width <- 3
} else
{
E(g)$width <- 5 * (E(g)$width-min(E(g)$width)) /
(max(E(g)$width)-min(E(g)$width)) + 0.5
}
}
# plot wTO network
layout(matrix(c(1, 2), 1, 2), c(2, 1), 1)
par(mar=c(5, 4, 4, 1.8))
image(matrix(spot.list$overview.mask, env$preferences$dim.1stLvlSom), col="gray90", axes=FALSE, main="Module correlations (WTO algorithm)", cex.main=1.6)
box()
par(new=TRUE, mar=c(3.9,2.7,2.7,0.5))
plot(g, layout=t(sapply(spot.list$spots, function(x) { x$position })),
vertex.label.color="black" ,vertex.label.cex=1, vertex.color="grey",
xlim=c(-1.0,1.0), ylim=c(-1.0,1.0))
par(mar=c(5, 1, 4, 2))
plot(0, type="n", axes=FALSE, xlab="", ylab="", xlim=c(0,1), ylim=c(0,1), xaxs="i", yaxs="i")
legend("center", c("negative correlation",expression(paste(" ",omega," < -0.25")),expression(paste(" ",omega," < -0.50")),expression(paste(" ",omega," < -0.75")),
"positive correlation",expression(paste(" ",omega," > 0.25")),expression(paste(" ",omega," > 0.50")),expression(paste(" ",omega," > 0.75"))),
lty = c(1,1), lwd = 4,col=c("white","rosybrown1","chocolate","brown4","white","lightblue","cornflowerblue","dodgerblue4"), bty="n")
box()
}
#### Baskets ####
if ( sum( rowSums(spotdata.binary) > 0 ) <=1 )
{
util.warn("Skipped basket and group association analyses due to missing module activation.")
return()
}
if ( length(spot.list$spots) > 2 )
{
module.group.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x)
{
samples.with.module <- names( which( spotdata.binary[mod.x,] == 1 ) )
do.call( rbind, lapply( unique( env$group.labels[ samples.with.module ] ), function(gr.x)
{
count <- sum( env$group.labels[ samples.with.module ] == gr.x )
data.frame( lhs = mod.x,
rhs = gr.x,
count = count,
support = count / ncol(env$indata),
confidence = count / length(samples.with.module),
stringsAsFactors = FALSE)
}) )
} ) )
module.group.rules <- module.group.rules[ which( module.group.rules$support>0.01 &
module.group.rules$confidence>0.1 ), ]
module.group.rules <- module.group.rules[ order( module.group.rules$confidence, decreasing=TRUE), ]
module.group.rules <- module.group.rules[ !duplicated( module.group.rules$lhs ) ,]
module.module.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x.1)
{
do.call( rbind, lapply( setdiff( rownames(spotdata.binary), mod.x.1), function(mod.x.2)
{
count <- sum( colSums( spotdata.binary[c(mod.x.1,mod.x.2),] ) == 2 )
data.frame( lhs = mod.x.1,
rhs = mod.x.2,
count = count,
support = count / ncol(env$indata),
confidence = count / sum( spotdata.binary[mod.x.1,] == 1 ),
stringsAsFactors = FALSE)
}) )
} ) )
module.module.rules <- module.module.rules[ which( module.module.rules$support>0.01 &
module.module.rules$confidence>0.1 ), ]
module.module.rules <- module.module.rules[ order( module.module.rules$confidence, decreasing=TRUE), ]
module.module.rules <- module.module.rules[ which( !duplicated( apply(cbind(module.module.rules$lhs,module.module.rules$rhs),1,function(x) paste(sort(x),collapse=" ") ) ) ),]
g <- graph.empty( length(spot.list$spots), directed = FALSE )
V(g)$name <- names(spot.list$spots)
g <- add_edges( g, as.vector( rbind( match( module.module.rules$lhs, V(g)$name ), match( module.module.rules$rhs, V(g)$name ) ) ) )
V(g)$label <- names(spot.list$spots)
V(g)$size <- 5
V(g)$size[ match( module.group.rules$lhs, V(g)$name ) ] <- 15 * module.group.rules$confidence + 3
V(g)$color <- "gray"
V(g)$color[ match( module.group.rules$lhs, V(g)$name ) ] <- env$groupwise.group.colors[module.group.rules$rhs]
E(g)$width <- 6 * module.module.rules$confidence + 0.5
E(g)$color <- "gray30"
E(g)$arrow.size <- 4
# plot implication network
layout(matrix(c(1, 2), 1, 2), c(2, 1), 1)
par(mar=c(5, 4, 4, 1.8))
image(matrix(spot.list$overview.mask, env$preferences$dim.1stLvlSom), col="gray90", axes=FALSE, main="Module implications (basket algorithm)", cex.main=1.6)
title("all samples",line=0.2)
box()
par(new=TRUE, mar=c(3.9,2.7,2.7,0.5))
plot(g, layout=t(sapply(spot.list$spots, function(x) { x$position })),
vertex.label.color="black" ,vertex.label.cex=1,
xlim=c(-1.0,1.0), ylim=c(-1.0,1.0))
par(mar=c(5, 1, 4, 2))
plot(0, type="n", axes=FALSE, xlab="", ylab="", xlim=c(0,1), ylim=c(0,1), xaxs="i", yaxs="i")
legend(0.08,0.95, c("Confidence","(group implication)","100%","50%","10%","","Confidence", "(module implication)","100%","50%","10%" ),
pch=c(NA,NA,1,1,1,rep(NA,6)), pt.cex=c(0,0,7.5,4,1,rep(0,6)), lty=c(rep(0,8),1,1,1), lwd=c(rep(0,8),7,4,1), bty="n", cex=1.2, y.intersp=2)
box()
for( gr in unique(env$group.labels) )
{
if( all( spotdata.binary[,which(env$group.labels==gr)] == 0 ) ) next
module.module.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x.1)
{
do.call( rbind, lapply( setdiff( rownames(spotdata.binary), mod.x.1), function(mod.x.2)
{
count <- sum( colSums( spotdata.binary[c(mod.x.1,mod.x.2),which(env$group.labels==gr),drop=FALSE] ) == 2 )
data.frame( lhs = mod.x.1,
rhs = mod.x.2,
count = count,
support = count / sum(env$group.labels==gr),
confidence = count / sum( spotdata.binary[mod.x.1,which(env$group.labels==gr)] == 1 ),
stringsAsFactors = FALSE)
}) )
} ) )
module.module.rules <- module.module.rules[ which( module.module.rules$support>0.1 &
module.module.rules$confidence>0.1 ), ]
module.module.rules <- module.module.rules[ order( module.module.rules$confidence, decreasing=TRUE), ]
module.module.rules <- module.module.rules[ which( !duplicated( apply(cbind(module.module.rules$lhs,module.module.rules$rhs),1,function(x) paste(sort(x),collapse=" ") ) ) ),]
if( nrow(module.module.rules) > 0 )
{
g <- graph.empty( length(spot.list$spots), directed = FALSE )
V(g)$name <- names(spot.list$spots)
g <- add_edges( g, as.vector( rbind( match( module.module.rules$lhs, V(g)$name ), match( module.module.rules$rhs, V(g)$name ) ) ) )
V(g)$label <- names(spot.list$spots)
V(g)$size <- 10
V(g)$color <- "gray"
V(g)$color[ match( module.group.rules$lhs[which(module.group.rules$rhs==gr)], V(g)$name ) ] <- env$groupwise.group.colors[gr]
E(g)$width <- 6 * module.module.rules$confidence + 0.5
E(g)$color <- env$groupwise.group.colors[gr]
layout(matrix(c(1, 2), 1, 2), c(2, 1), 1)
par(mar=c(5, 4, 4, 1.8))
image(matrix(spot.list$overview.mask, env$preferences$dim.1stLvlSom), col="gray90", axes=FALSE, main="Module implications (basket algorithm)", cex.main=1.6)
title(paste("Group samples:",gr),line=0.2,col.main=env$groupwise.group.colors[gr])
box()
par(new=TRUE, mar=c(3.9,2.7,2.7,0.5))
plot(g, layout=t(sapply(spot.list$spots, function(x) { x$position })),
vertex.label.color="black" ,vertex.label.cex=1,
xlim=c(-1.0,1.0), ylim=c(-1.0,1.0))
par(mar=c(5, 1, 4, 2))
plot(0, type="n", axes=FALSE, xlab="", ylab="", xlim=c(0,1), ylim=c(0,1), xaxs="i", yaxs="i")
legend(0.08,0.95, c("Confidence","(module implication)","100%","50%","10%" ),
lty=c(rep(0,2),1,1,1), lwd=c(rep(0,2),7,4,1), bty="n", cex=1.2, y.intersp=2)
box()
}
}
}
#### Group Associations ####
if (length(unique(env$group.labels)) > 1)
{
layout( matrix(c(rep(1,4),2:13),nrow=4,byrow=TRUE), heights=c(0.15,1,1,1) )
par(mar=c(0,0,0,0))
frame()
text(0.5,0.5,"Group associations of the modules",cex=2.6)
i <- 1
for( rowname in rownames(spotdata.binary) )
{
if(i>1&&i%%13==0) { par(mar=c(0,0,0,0)); frame(); i <- i + 1 }
count <- tapply(spotdata.binary[rowname,],env$group.labels,sum)[unique(env$group.labels)]
percent <- count/table(env$group.labels)[unique(env$group.labels)]
count <- count[which(count>0)]
percent <- percent[which(percent>0)]
if( length(percent) > 0 )
{
par(mar=c(5,6,4,5))
barplot( 100*percent, horiz=TRUE, main=paste(rowname," (",sum(count),")",sep=""), cex.main=1.6, xlim=c(0,100),
names.arg=paste(names(percent),"\n(",count,")",sep="" ), las=1, xlab="association (%)",
col=env$groupwise.group.colors[names(percent)] )
i <- i + 1
}
}
}
#### Group Implications ####
if ( length(spot.list$spots) > 2 )
{
group.module.rules <-
do.call( rbind, lapply( unique( env$group.labels ), function(gr.x)
{
samples.with.group <- names( which( env$group.labels == gr.x ) )
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x)
{
count <- sum( spotdata.binary[mod.x,samples.with.group ] == 1 )
data.frame( lhs = gr.x,
rhs = mod.x,
count = count,
support = count / ncol(env$indata),
confidence = count / length(samples.with.group),
stringsAsFactors = FALSE)
}) )
} ) )
group.module.rules <- group.module.rules[ which( group.module.rules$support>0.01 &
group.module.rules$confidence>0.01 ), ]
group.module.rules <- group.module.rules[ order( group.module.rules$confidence), ]
layout( matrix(c(rep(1,4),2:13),nrow=4,byrow=TRUE), heights=c(0.15,1,1,1) )
par(mar=c(0,0,0,0))
frame()
text(0.5,0.5,"Module implications of the groups (basket algorithm)",cex=2.6)
i <- 1
for( gr in unique(env$group.labels) )
{
if(i>1&&i%%13==0) { par(mar=c(0,0,0,0)); frame(); i <- i + 1 }
r <- which( group.module.rules$lhs == gr )
if( length(r) > 0 )
{
par(mar=c(5,6,4,5))
barplot( 100*group.module.rules[r,]$confidence, horiz=TRUE, main=paste(gr," (",sum(env$group.labels==gr),")",sep=""), col.main=env$groupwise.group.colors[gr], cex.main=1.6, xlim=c(0,100),
names.arg=paste(group.module.rules[r,]$rhs,"\n(",group.module.rules[r,]$count,")",sep="" ), las=1, xlab="confidence" )
title(main="implies",line=0,cex.main=0.8)
i <- i + 1
}
}
module.group.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x)
{
samples.with.module <- names( which( spotdata.binary[mod.x,] == 1 ) )
do.call( rbind, lapply( unique( env$group.labels[ samples.with.module ] ), function(gr.x)
{
count <- sum( env$group.labels[ samples.with.module ] == gr.x )
data.frame( lhs = mod.x,
rhs = gr.x,
count = count,
support = count / ncol(env$indata),
confidence = count / length(samples.with.module),
stringsAsFactors = FALSE)
}) )
} ) )
module.group.rules <- module.group.rules[ which( module.group.rules$support>0.01 &
module.group.rules$confidence>0.01 ), ]
module.group.rules <- module.group.rules[ order( module.group.rules$confidence ), ]
layout( matrix(c(rep(1,4),2:13),nrow=4,byrow=TRUE), heights=c(0.15,1,1,1) )
par(mar=c(0,0,0,0))
frame()
text(0.5,0.5,"Group implications of the modules (basket algorithm)",cex=2.6)
i <- 1
for( lhs in unique(rev(module.group.rules$lhs) ) )
{
if(i>1&&i%%13==0) { par(mar=c(0,0,0,0)); frame(); i <- i + 1 }
r <- which( module.group.rules$lhs == lhs )
n <- module.group.rules[r[1],]$count / module.group.rules[r[1],]$confidence
if( length(r) > 0 )
{
par(mar=c(5,8,4,5))
barplot( 100*module.group.rules[r,]$confidence, horiz=TRUE, main=paste(lhs," (",n,")",sep=""), cex.main=1.6, xlim=c(0,100),
names.arg=paste(module.group.rules[r,]$rhs,"\n(",module.group.rules[r,]$count,")",sep="" ), las=1, xlab="confidence",
col=env$groupwise.group.colors[module.group.rules[r,]$rhs] )
title(main="implies",line=0,cex.main=0.8)
i <- i + 1
}
}
}
dev.off()
}
hloefflerwirth/oposSOM documentation built on Oct. 29, 2024, 4:12 a.m.
R Package Documentation
Browse R Packages
We want your feedback!
Note that we can't provide technical support on individual packages. You should contact the package authors for that.
Defines functions modules.relations
modules.relations <- function(env, spot.list, main, path)
{
sd.theshold <- sd(spot.list$spotdata)
spotdata.binary <- spot.list$spotdata > sd.theshold
mode(spotdata.binary) <- "numeric"
pdf(path, 29.7/2.54, 21/2.54, useDingbats=FALSE)
#### wTO network ####
if ( length(spot.list$spots) > 2 )
{
# calculate wTO #
adj_matrix <- cor(t(spot.list$spotdata),t(spot.list$spotdata))
diag(adj_matrix) <- 0
omega <- matrix(NA,
nrow=nrow(spot.list$spotdata),
ncol=nrow(spot.list$spotdata),
dimnames=list(names(spot.list$spots),names(spot.list$spots)))
b <- adj_matrix %*% adj_matrix
k_sum <- apply(adj_matrix, 2, function(x) { sum(abs(x)) })
for (i in 1: nrow(spot.list$spotdata))
{
for (j in 1: nrow(spot.list$spotdata))
{
omega[i,j] <-
(b[i,j] + adj_matrix[i,j]) /
(min(k_sum[i],k_sum[j]) + 1 - abs(adj_matrix[i,j]))
}
}
diag(omega) <- 0
omega[which(abs(omega) < 0.25)] <- 0
g <- graph.adjacency(omega, weighted=TRUE, mode="undirected")
V(g)$label <- names(spot.list$spots)
n.spot <- apply(spot.list$spotdata, 1, function(x) sum(x>sd.theshold) )
if (max(n.spot) == min(n.spot))
{
V(g)$size <- 10
} else
{
V(g)$size <- 8 * (n.spot-min(n.spot)) / (max(n.spot)-min(n.spot)) + 6
}
if (length(E(g)) > 0)
{
E(g)$color <- c("brown4","chocolate","rosybrown1","white","white","lightblue","cornflowerblue","dodgerblue4")[ cut( E(g)$weight, breaks=c(-Inf,-0.75,-0.5,-0.25,0,0.25,0.5,0.75,Inf), labels=c(1:8) ) ]
E(g)$width <- abs(E(g)$weight)
if (max(E(g)$width) == min(E(g)$width))
{
E(g)$width <- 3
} else
{
E(g)$width <- 5 * (E(g)$width-min(E(g)$width)) /
(max(E(g)$width)-min(E(g)$width)) + 0.5
}
}
# plot wTO network
layout(matrix(c(1, 2), 1, 2), c(2, 1), 1)
par(mar=c(5, 4, 4, 1.8))
image(matrix(spot.list$overview.mask, env$preferences$dim.1stLvlSom), col="gray90", axes=FALSE, main="Module correlations (WTO algorithm)", cex.main=1.6)
box()
par(new=TRUE, mar=c(3.9,2.7,2.7,0.5))
plot(g, layout=t(sapply(spot.list$spots, function(x) { x$position })),
vertex.label.color="black" ,vertex.label.cex=1, vertex.color="grey",
xlim=c(-1.0,1.0), ylim=c(-1.0,1.0))
par(mar=c(5, 1, 4, 2))
plot(0, type="n", axes=FALSE, xlab="", ylab="", xlim=c(0,1), ylim=c(0,1), xaxs="i", yaxs="i")
legend("center", c("negative correlation",expression(paste(" ",omega," < -0.25")),expression(paste(" ",omega," < -0.50")),expression(paste(" ",omega," < -0.75")),
"positive correlation",expression(paste(" ",omega," > 0.25")),expression(paste(" ",omega," > 0.50")),expression(paste(" ",omega," > 0.75"))),
lty = c(1,1), lwd = 4,col=c("white","rosybrown1","chocolate","brown4","white","lightblue","cornflowerblue","dodgerblue4"), bty="n")
box()
}
#### Baskets ####
if ( sum( rowSums(spotdata.binary) > 0 ) <=1 )
{
util.warn("Skipped basket and group association analyses due to missing module activation.")
return()
}
if ( length(spot.list$spots) > 2 )
{
module.group.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x)
{
samples.with.module <- names( which( spotdata.binary[mod.x,] == 1 ) )
do.call( rbind, lapply( unique( env$group.labels[ samples.with.module ] ), function(gr.x)
{
count <- sum( env$group.labels[ samples.with.module ] == gr.x )
data.frame( lhs = mod.x,
rhs = gr.x,
count = count,
support = count / ncol(env$indata),
confidence = count / length(samples.with.module),
stringsAsFactors = FALSE)
}) )
} ) )
module.group.rules <- module.group.rules[ which( module.group.rules$support>0.01 &
module.group.rules$confidence>0.1 ), ]
module.group.rules <- module.group.rules[ order( module.group.rules$confidence, decreasing=TRUE), ]
module.group.rules <- module.group.rules[ !duplicated( module.group.rules$lhs ) ,]
module.module.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x.1)
{
do.call( rbind, lapply( setdiff( rownames(spotdata.binary), mod.x.1), function(mod.x.2)
{
count <- sum( colSums( spotdata.binary[c(mod.x.1,mod.x.2),] ) == 2 )
data.frame( lhs = mod.x.1,
rhs = mod.x.2,
count = count,
support = count / ncol(env$indata),
confidence = count / sum( spotdata.binary[mod.x.1,] == 1 ),
stringsAsFactors = FALSE)
}) )
} ) )
module.module.rules <- module.module.rules[ which( module.module.rules$support>0.01 &
module.module.rules$confidence>0.1 ), ]
module.module.rules <- module.module.rules[ order( module.module.rules$confidence, decreasing=TRUE), ]
module.module.rules <- module.module.rules[ which( !duplicated( apply(cbind(module.module.rules$lhs,module.module.rules$rhs),1,function(x) paste(sort(x),collapse=" ") ) ) ),]
g <- graph.empty( length(spot.list$spots), directed = FALSE )
V(g)$name <- names(spot.list$spots)
g <- add_edges( g, as.vector( rbind( match( module.module.rules$lhs, V(g)$name ), match( module.module.rules$rhs, V(g)$name ) ) ) )
V(g)$label <- names(spot.list$spots)
V(g)$size <- 5
V(g)$size[ match( module.group.rules$lhs, V(g)$name ) ] <- 15 * module.group.rules$confidence + 3
V(g)$color <- "gray"
V(g)$color[ match( module.group.rules$lhs, V(g)$name ) ] <- env$groupwise.group.colors[module.group.rules$rhs]
E(g)$width <- 6 * module.module.rules$confidence + 0.5
E(g)$color <- "gray30"
E(g)$arrow.size <- 4
# plot implication network
layout(matrix(c(1, 2), 1, 2), c(2, 1), 1)
par(mar=c(5, 4, 4, 1.8))
image(matrix(spot.list$overview.mask, env$preferences$dim.1stLvlSom), col="gray90", axes=FALSE, main="Module implications (basket algorithm)", cex.main=1.6)
title("all samples",line=0.2)
box()
par(new=TRUE, mar=c(3.9,2.7,2.7,0.5))
plot(g, layout=t(sapply(spot.list$spots, function(x) { x$position })),
vertex.label.color="black" ,vertex.label.cex=1,
xlim=c(-1.0,1.0), ylim=c(-1.0,1.0))
par(mar=c(5, 1, 4, 2))
plot(0, type="n", axes=FALSE, xlab="", ylab="", xlim=c(0,1), ylim=c(0,1), xaxs="i", yaxs="i")
legend(0.08,0.95, c("Confidence","(group implication)","100%","50%","10%","","Confidence", "(module implication)","100%","50%","10%" ),
pch=c(NA,NA,1,1,1,rep(NA,6)), pt.cex=c(0,0,7.5,4,1,rep(0,6)), lty=c(rep(0,8),1,1,1), lwd=c(rep(0,8),7,4,1), bty="n", cex=1.2, y.intersp=2)
box()
for( gr in unique(env$group.labels) )
{
if( all( spotdata.binary[,which(env$group.labels==gr)] == 0 ) ) next
module.module.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x.1)
{
do.call( rbind, lapply( setdiff( rownames(spotdata.binary), mod.x.1), function(mod.x.2)
{
count <- sum( colSums( spotdata.binary[c(mod.x.1,mod.x.2),which(env$group.labels==gr),drop=FALSE] ) == 2 )
data.frame( lhs = mod.x.1,
rhs = mod.x.2,
count = count,
support = count / sum(env$group.labels==gr),
confidence = count / sum( spotdata.binary[mod.x.1,which(env$group.labels==gr)] == 1 ),
stringsAsFactors = FALSE)
}) )
} ) )
module.module.rules <- module.module.rules[ which( module.module.rules$support>0.1 &
module.module.rules$confidence>0.1 ), ]
module.module.rules <- module.module.rules[ order( module.module.rules$confidence, decreasing=TRUE), ]
module.module.rules <- module.module.rules[ which( !duplicated( apply(cbind(module.module.rules$lhs,module.module.rules$rhs),1,function(x) paste(sort(x),collapse=" ") ) ) ),]
if( nrow(module.module.rules) > 0 )
{
g <- graph.empty( length(spot.list$spots), directed = FALSE )
V(g)$name <- names(spot.list$spots)
g <- add_edges( g, as.vector( rbind( match( module.module.rules$lhs, V(g)$name ), match( module.module.rules$rhs, V(g)$name ) ) ) )
V(g)$label <- names(spot.list$spots)
V(g)$size <- 10
V(g)$color <- "gray"
V(g)$color[ match( module.group.rules$lhs[which(module.group.rules$rhs==gr)], V(g)$name ) ] <- env$groupwise.group.colors[gr]
E(g)$width <- 6 * module.module.rules$confidence + 0.5
E(g)$color <- env$groupwise.group.colors[gr]
layout(matrix(c(1, 2), 1, 2), c(2, 1), 1)
par(mar=c(5, 4, 4, 1.8))
image(matrix(spot.list$overview.mask, env$preferences$dim.1stLvlSom), col="gray90", axes=FALSE, main="Module implications (basket algorithm)", cex.main=1.6)
title(paste("Group samples:",gr),line=0.2,col.main=env$groupwise.group.colors[gr])
box()
par(new=TRUE, mar=c(3.9,2.7,2.7,0.5))
plot(g, layout=t(sapply(spot.list$spots, function(x) { x$position })),
vertex.label.color="black" ,vertex.label.cex=1,
xlim=c(-1.0,1.0), ylim=c(-1.0,1.0))
par(mar=c(5, 1, 4, 2))
plot(0, type="n", axes=FALSE, xlab="", ylab="", xlim=c(0,1), ylim=c(0,1), xaxs="i", yaxs="i")
legend(0.08,0.95, c("Confidence","(module implication)","100%","50%","10%" ),
lty=c(rep(0,2),1,1,1), lwd=c(rep(0,2),7,4,1), bty="n", cex=1.2, y.intersp=2)
box()
}
}
}
#### Group Associations ####
if (length(unique(env$group.labels)) > 1)
{
layout( matrix(c(rep(1,4),2:13),nrow=4,byrow=TRUE), heights=c(0.15,1,1,1) )
par(mar=c(0,0,0,0))
frame()
text(0.5,0.5,"Group associations of the modules",cex=2.6)
i <- 1
for( rowname in rownames(spotdata.binary) )
{
if(i>1&&i%%13==0) { par(mar=c(0,0,0,0)); frame(); i <- i + 1 }
count <- tapply(spotdata.binary[rowname,],env$group.labels,sum)[unique(env$group.labels)]
percent <- count/table(env$group.labels)[unique(env$group.labels)]
count <- count[which(count>0)]
percent <- percent[which(percent>0)]
if( length(percent) > 0 )
{
par(mar=c(5,6,4,5))
barplot( 100*percent, horiz=TRUE, main=paste(rowname," (",sum(count),")",sep=""), cex.main=1.6, xlim=c(0,100),
names.arg=paste(names(percent),"\n(",count,")",sep="" ), las=1, xlab="association (%)",
col=env$groupwise.group.colors[names(percent)] )
i <- i + 1
}
}
}
#### Group Implications ####
if ( length(spot.list$spots) > 2 )
{
group.module.rules <-
do.call( rbind, lapply( unique( env$group.labels ), function(gr.x)
{
samples.with.group <- names( which( env$group.labels == gr.x ) )
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x)
{
count <- sum( spotdata.binary[mod.x,samples.with.group ] == 1 )
data.frame( lhs = gr.x,
rhs = mod.x,
count = count,
support = count / ncol(env$indata),
confidence = count / length(samples.with.group),
stringsAsFactors = FALSE)
}) )
} ) )
group.module.rules <- group.module.rules[ which( group.module.rules$support>0.01 &
group.module.rules$confidence>0.01 ), ]
group.module.rules <- group.module.rules[ order( group.module.rules$confidence), ]
layout( matrix(c(rep(1,4),2:13),nrow=4,byrow=TRUE), heights=c(0.15,1,1,1) )
par(mar=c(0,0,0,0))
frame()
text(0.5,0.5,"Module implications of the groups (basket algorithm)",cex=2.6)
i <- 1
for( gr in unique(env$group.labels) )
{
if(i>1&&i%%13==0) { par(mar=c(0,0,0,0)); frame(); i <- i + 1 }
r <- which( group.module.rules$lhs == gr )
if( length(r) > 0 )
{
par(mar=c(5,6,4,5))
barplot( 100*group.module.rules[r,]$confidence, horiz=TRUE, main=paste(gr," (",sum(env$group.labels==gr),")",sep=""), col.main=env$groupwise.group.colors[gr], cex.main=1.6, xlim=c(0,100),
names.arg=paste(group.module.rules[r,]$rhs,"\n(",group.module.rules[r,]$count,")",sep="" ), las=1, xlab="confidence" )
title(main="implies",line=0,cex.main=0.8)
i <- i + 1
}
}
module.group.rules <-
do.call( rbind, lapply( rownames(spotdata.binary), function(mod.x)
{
samples.with.module <- names( which( spotdata.binary[mod.x,] == 1 ) )
do.call( rbind, lapply( unique( env$group.labels[ samples.with.module ] ), function(gr.x)
{
count <- sum( env$group.labels[ samples.with.module ] == gr.x )
data.frame( lhs = mod.x,
rhs = gr.x,
count = count,
support = count / ncol(env$indata),
confidence = count / length(samples.with.module),
stringsAsFactors = FALSE)
}) )
} ) )
module.group.rules <- module.group.rules[ which( module.group.rules$support>0.01 &
module.group.rules$confidence>0.01 ), ]
module.group.rules <- module.group.rules[ order( module.group.rules$confidence ), ]
layout( matrix(c(rep(1,4),2:13),nrow=4,byrow=TRUE), heights=c(0.15,1,1,1) )
par(mar=c(0,0,0,0))
frame()
text(0.5,0.5,"Group implications of the modules (basket algorithm)",cex=2.6)
i <- 1
for( lhs in unique(rev(module.group.rules$lhs) ) )
{
if(i>1&&i%%13==0) { par(mar=c(0,0,0,0)); frame(); i <- i + 1 }
r <- which( module.group.rules$lhs == lhs )
n <- module.group.rules[r[1],]$count / module.group.rules[r[1],]$confidence
if( length(r) > 0 )
{
par(mar=c(5,8,4,5))
barplot( 100*module.group.rules[r,]$confidence, horiz=TRUE, main=paste(lhs," (",n,")",sep=""), cex.main=1.6, xlim=c(0,100),
names.arg=paste(module.group.rules[r,]$rhs,"\n(",module.group.rules[r,]$count,")",sep="" ), las=1, xlab="confidence",
col=env$groupwise.group.colors[module.group.rules[r,]$rhs] )
title(main="implies",line=0,cex.main=0.8)
i <- i + 1
}
}
}
dev.off()
}
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.