minmax: Minmax scaling

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

Description

Perform a minmax standardisation to scale data into 0 to 1 range

Usage

minmax(mat)

Arguments

mat	a matrix with rows correspond to phosphosites and columns correspond to condition

Value

Minmax standardised matrix

Examples

data('phospho_L6_ratio')
data('SPSs')

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

# Cleaning phosphosite label
phospho.site.names = rownames(phospho.L6.ratio)
L6.sites = gsub(' ', '', sapply(strsplit(rownames(phospho.L6.ratio), '~'),
                                function(x){paste(toupper(x[2]), x[3], '',
                                                sep=';')}))
phospho.L6.ratio = t(sapply(split(data.frame(phospho.L6.ratio), L6.sites),
                            colMeans))
phospho.site.names = split(phospho.site.names, L6.sites)

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

# phosphoproteomics data normalisation using RUV
ctl = which(rownames(phospho.L6.ratio) %in% SPSs)
phospho.L6.ratio.RUV = RUVphospho(phospho.L6.ratio, M = design, k = 3,
                                ctl = ctl)

phosphoL6 = phospho.L6.ratio.RUV
rownames(phosphoL6) = phospho.site.names

# filter for up-regulated phosphosites
phosphoL6.mean <- meanAbundance(phosphoL6, grps = gsub('_.+', '',
                                colnames(phosphoL6)))
aov <- matANOVA(mat=phosphoL6, grps=gsub('_.+', '', colnames(phosphoL6)))
phosphoL6.reg <- phosphoL6[(aov < 0.05) &
                        (rowSums(phosphoL6.mean > 0.5) > 0),,drop = FALSE]
L6.phos.std <- standardise(phosphoL6.reg)
rownames(L6.phos.std) <- sapply(strsplit(rownames(L6.phos.std), '~'),
    function(x){gsub(' ', '', paste(toupper(x[2]), x[3], '', sep=';'))})

L6.phos.seq <- sapply(strsplit(rownames(phosphoL6.reg), '~'),
                    function(x)x[4])

numMotif = 5
numSub = 1

ks.profile.list <- kinaseSubstrateProfile(PhosphoSite.mouse, L6.phos.std)
motif.mouse.list = PhosR::motif.mouse.list

motif.mouse.list.filtered <-
    motif.mouse.list[which(motif.mouse.list$NumInputSeq >= numMotif)]
ks.profile.list.filtered <-
    ks.profile.list[which(ks.profile.list$NumSub >= numSub)]

# scoring all phosphosites against all motifs
motifScoreMatrix <-
    matrix(NA, nrow=nrow(L6.phos.std),
    ncol=length(motif.mouse.list.filtered))
rownames(motifScoreMatrix) <- rownames(L6.phos.std)
colnames(motifScoreMatrix) <- names(motif.mouse.list.filtered)

# extracting flanking sequences
seqWin = mapply(function(x) {
    mid <- (nchar(x)+1)/2
    substr(x, start=(mid-7), stop=(mid+7))
}, L6.phos.seq)


print('Scoring phosphosites against kinase motifs:')
for(i in seq_len(length(motif.mouse.list.filtered))) {
    motifScoreMatrix[,i] <-
        frequencyScoring(seqWin, motif.mouse.list.filtered[[i]])
        cat(paste(i, '.', sep=''))
}
motifScoreMatrix <- minmax(motifScoreMatrix)

PengyiYang/PhosR documentation built on June 21, 2020, 8:37 a.m.