phosCollapse: Summarising phosphosites to proteins

In PYangLab/PhosR: A set of methods and tools for comprehensive analysis of phosphoproteomics data

Summarising phosphosites to proteins

Description

Summarising phosphosite-level information to proteins for performing downstream gene-centric analyses.

Usage

phosCollapse(mat, id, stat, by='min')

Arguments

mat	a matrix with rows correspond to phosphosites and columns correspond to samples.
id	an array indicating the groupping of phosphosites etc.
stat	an array containing statistics of phosphosite such as phosphorylation levels.
by	how to summarise phosphosites using their statistics. Either by 'min' (default), 'max', or 'mid'.

Value

A matrix summarised to protein level

Examples

library(limma)

data('phospho_L6_ratio_pe')
data('SPSs')

grps = gsub('_.+', '', colnames(phospho.L6.ratio.pe))

L6.sites = paste(sapply(GeneSymbol(phospho.L6.ratio.pe), function(x)paste(x)),
                 ";",
                 sapply(Residue(phospho.L6.ratio.pe), function(x)paste(x)),
                 sapply(Site(phospho.L6.ratio.pe), function(x)paste(x)),
                 ";", sep = "")

# Construct a design matrix by condition
design = model.matrix(~ grps - 1)

ctl = which(L6.sites %in% SPSs)
phospho.L6.ratio.pe = RUVphospho(phospho.L6.ratio.pe, 
                                  M = design, k = 3, ctl = ctl)

# fit linear model for each phosphosite
f <- grps
X <- model.matrix(~ f - 1)
fit <- lmFit(SummarizedExperiment::assay(phospho.L6.ratio.pe, "normalised"), X)

# extract top-ranked phosphosites for each condition compared to basal
table.AICAR <- topTable(eBayes(fit), number=Inf, coef = 1)
table.Ins <- topTable(eBayes(fit), number=Inf, coef = 3)
table.AICARIns <- topTable(eBayes(fit), number=Inf, coef = 2)

DE1.RUV <- c(sum(table.AICAR[,'adj.P.Val'] < 0.05),
    sum(table.Ins[,'adj.P.Val'] < 0.05),
    sum(table.AICARIns[,'adj.P.Val'] < 0.05))

# extract top-ranked phosphosites for each group comparison
contrast.matrix1 <- makeContrasts(fAICARIns-fIns, levels=X)
contrast.matrix2 <- makeContrasts(fAICARIns-fAICAR, levels=X)
fit1 <- contrasts.fit(fit, contrast.matrix1)
fit2 <- contrasts.fit(fit, contrast.matrix2)
table.AICARInsVSIns <- topTable(eBayes(fit1), number=Inf)
table.AICARInsVSAICAR <- topTable(eBayes(fit2), number=Inf)

DE2.RUV <- c(sum(table.AICARInsVSIns[,'adj.P.Val'] < 0.05),
    sum(table.AICARInsVSAICAR[,'adj.P.Val'] < 0.05))

o <- rownames(table.AICARInsVSIns)
Tc <- cbind(table.Ins[o,'logFC'], table.AICAR[o,'logFC'],
            table.AICARIns[o,'logFC'])
rownames(Tc) = gsub('(.*)(;[A-Z])([0-9]+)(;)', '\\1;\\3;', o)
colnames(Tc) <- c('Ins', 'AICAR', 'AICAR+Ins')

# summary phosphosite-level information to proteins for performing downstream
# gene-centric analyses.
Tc.gene <- phosCollapse(Tc, id=gsub(';.+', '', rownames(Tc)),
    stat=apply(abs(Tc), 1, max), by = 'max')

PYangLab/PhosR documentation built on Nov. 1, 2024, 1:47 p.m.