suppressPackageStartupMessages(library("synapter")) suppressPackageStartupMessages(library("synapterdata")) suppressPackageStartupMessages(library("BiocStyle")) synobj2RData()
This document assumes familiarity with standard r Biocpkg("synapter")
pipeline described in [@Bond2013] and in the package
r Biocpkg("synapter")
vignette, available
online
and with vignette("synapter", package = "synapter")
.
In this vignette we introduce a new fragment matching feature (see figures
2, 3 and 4) which
improves the matching of identification and the quantitation features. After
applying the usual synergise1
workflow (see ?synergise1
and ?Synapter
for details) a number of multiple matches and possible false unique matches
remain that can be deconvoluted by comparing common peaks in the identification
fragment peaks and the quantitation spectra.
The example data synobj2
used throughout this document is available in the
r Biocpkg("synapterdata")
package and can be directly load as follows:
library("synapterdata") synobj2RData()
In the [next section](\sec:synergise] we describe how synobj2
was generated.
The test files used in this vignette can be downloaded from
http://proteome.sysbiol.cam.ac.uk/lgatto/synapter/data/.
The following sections then describe the new fragment matching functionality.
synergise1
{#sec:synergise}One has to run the synergise1
workflow before fragment
matching can be applied. Please read the general r Biocpkg("synapter")
vignette
for the general use of synergise1
.
The additional data needed for the fragment matching procedure are a
final_fragment.csv
file for the identification run and a Spectrum.xml
file
for the quantitation run.
cat(readLines(system.file(file.path("scripts", "create_synobj2.R"), package="synapterdata"), n=13), sep="\n")
synobj2 <- synergise1(object=synobj2, outputdir=tempdir())
This step is optional and allows one to remove low abundance fragments
in the spectra using filterFragments
. Filtering fragments can remove noise
in the spectra and reduce undesired fragment matches. Prior to filtering, the
plotCumulativeNumberOfFragments
function can be use to visualise the
intensity of all fragments. Both functions have an argument what
to decide
what spectra/fragments to filter/plot. Choose fragments.ident
for the
identification fragments and spectra.quant
for the quantitation fragments.
plotCumulativeNumberOfFragments(synobj2, what = "fragments.ident") plotCumulativeNumberOfFragments(synobj2, what = "spectra.quant")
filterFragments(synobj2, what = "fragments.ident", minIntensity = 70) filterFragments(synobj2, what = "spectra.quant", minIntensity = 70)
This method, named fragmentMatching
, performs the matching of the
identification fragments vs the quantitation spectra and counts the number
of identical peaks for each combination.
Because the peaks/fragments in the spectra of one run will never be
numerically identical to these in another, a tolerance parameter has
to be set using the setFragmentMatchingPpmTolerance
function.
Peaks/Fragments within this tolerance are treated as identical.
setFragmentMatchingPpmTolerance(synobj2, 25) fragmentMatching(synobj2)
The plotFragmentMatching
function illustrates the details of this fragment
matching procedure. If it is called without any additional argument every
matched pair (fragment vs spectrum) is plotted.
One can use the key
argument to select a special value in a column
(defined by the column
argument) of the MatchedEMRTs
data.frame
.
E.g. if one wants to select the fragment matching results with a high number
of common peaks, e.g. 28 common peaks:
plotFragmentMatching(synobj2, key = 28, column = "FragmentMatching")
Or, if one is interested in all results for the peptide with the sequence
"TALIDGLAQR"
.
plotFragmentMatching(synobj2, key = "TALIDGLAQR", column = "peptide.seq")
Maybe the peptide with a special precursor ID looks interesting.
plotFragmentMatching(synobj2, key = 12589, column = "precursor.leID.ident")
The plotFragmentMatchingPerformance
function can be used to
assess the performance of the fragment matching and the result of the
filtering procedure (see below) based on the number of common
peaks. This function invisibly returns a list with matrices containing
true positive, false positive, true negative and false negative
matches for the unique and non unique matches EMRT matches, as
illustrated in tables 1 and
2.
Both tables could be also generated by fragmentMatchingPerformance
.
m <- plotFragmentMatchingPerformance(synobj2)
knitr::kable(m$unique[1:15,], row.names=FALSE, caption="Number of true positives, false negatives, false positives, false negatives and false discovery rate for a given number of common peaks.")
knitr::kable(m$nonunique[1:15,], row.names=FALSE, caption="Number of true positives, false negatives, false positives, false negatives and false discovery rate for a given difference in number of common peaks between the higest and second highest multiply matching EMRTs in terms of number of common peptides.")
From the left panel on figure 7 and table 1 displaying counts for unique matches one can define filtering values for the unique (this section) and multiple matches (next section). In the case of uniquely matching EMRTs, one can easily reduce the number of false matches by requiring that true matches must have at least one peak/fragment in common. Clearly this will also remove some true matches. The question is whether you want to rely on matches that have no (or only a few) peaks/fragments in common?
performance(synobj2) getEMRTtable(synobj2) filterUniqueMatches(synobj2, minNumber = 1) performance(synobj2) getEMRTtable(synobj2)
The largest benefit of fragment matching is for non unique matches. If we assume that true match have a highest number of common peaks/fragments, we can distinguish correct matches among multiple possible matches that could not resolved before (c.f. section [#sec:plotperf]). To do so, we use the difference of common peaks from the highest to the second highest number in the match group. Assuming two cases with multiple matches. In the first case, we have two possible matches: a match with 7 and a match with 2 fragments in common. In the second ambiguous match, there are a matches with 2 and 1 fragments in common respectively. If we decide to accept a difference of at least 2, our first multiple match case be resolved into a unique match as the difference between the best and second matches is 5 and the best match with 7 common fragments will be upgraded to a unique match.
The right panel of figure 7 and table 2 can be used to choose a good threshold for the difference in number of common peaks.
oldEMRTtable <- getEMRTtable(synobj2) performance(synobj2) getEMRTtable(synobj2) filterNonUniqueMatches(synobj2, minDelta = 2) performance(synobj2) getEMRTtable(synobj2)
In this example we rescued r getEMRTtable(synobj2)["1"]-oldEMRTtable["1"]
unique
matches out of the non unique ones.
Like in the initial r Biocpkg("synapter")
workflow, it is possible to export the
MatchedEMRT
results using the writeMatchedEMRTs
function. The table has
some new columns that correspond to the fragment matching procedure,
e.g. FragmentMatching
, \dots.
writeMatchedEMRTs(synobj2, file = "MatchedEMRTs.csv")
All software and respective versions used to produce this document are listed below.
sessionInfo()
Any scripts or data that you put into this service are public.
Add the following code to your website.
For more information on customizing the embed code, read Embedding Snippets.