kinaseSubstratePred {PhosR} | R Documentation |
A machine learning approach for predicting specific kinase for a given substrate. This prediction framework utilise adaptive sampling.
kinaseSubstratePred( phosScoringMatrices, ensembleSize = 10, top = 50, cs = 0.8, inclusion = 20, iter = 5 )
phosScoringMatrices |
An output of kinaseSubstrateScore. |
ensembleSize |
An ensemble size. |
top |
a number to select top kinase substrates. |
cs |
Score threshold. |
inclusion |
A minimal number of substrates required for a kinase to be selected. |
iter |
A number of iterations for adaSampling. |
Kinase prediction matrix
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]) L6.matrices <- kinaseSubstrateScore(PhosphoSite.mouse, L6.phos.std, L6.phos.seq, numMotif = 5, numSub = 1) set.seed(1) L6.predMat <- kinaseSubstratePred(L6.matrices, top=30)