nQuant/featureCV return wrong sized matrix

[sgibb]

This PR fixes nQuant and featureCV. Both functions return a wrong-sized matrix if the grouping variable is a factor that contains more levels than items:
MWE:

library("MSnbase")

m <- new("MSnSet",
         exprs=matrix(1:10, nrow=5, ncol=2),
         featureData=new("AnnotatedDataFrame",
                         data=data.frame(accession=
                                         factor(c("A", "A", "A", "B", "B"),
                                                levels=LETTERS[1:10]))))
exprs(m)
#   1  2
# 1 1  6
# 2 2  7
# 3 3  8
# 4 4  9
# 5 5 10

nQuants(m, fData(m)$accession)
#   1 2
# A 3 3
# B 2 2
# C 0 0
# D 0 0
# E 0 0
# F 0 0
# G 0 0
# H 0 0
# I 0 0
# J 0 0

This results in an error if you want to apply TOP3 quantitation after you did some subsetting (e.g. filtering):

library("MSnbase")
data(msnset)

nQuants(msnset, fData(msnset)$ProteinAccession)
#         iTRAQ4.114 iTRAQ4.115 iTRAQ4.116 iTRAQ4.117
# BSA              3          3          3          3
# ECA0172          1          1          1          1
# ECA0435          2          2          2          2
# ...
# ECA4514          6          6          6          6
# ENO              4          4          3          4

msnsetSubset <- msnset[1:5]
nQuants(msnsetSubset, fData(msnsetSubset)$ProteinAccession)
#         iTRAQ4.114 iTRAQ4.115 iTRAQ4.116 iTRAQ4.117
# BSA              1          1          1          1
# ECA0172          0          0          0          0
# ECA0435          0          0          0          0
# ...
# ECA4514          0          0          0          0
# ENO              0          0          0          0


msnsetSubsetTop3 <- topN(msnsetSubset,
                         groupBy=fData(msnsetSubset)$ProteinAccession,
                         n=3)
nSubsetPeps <- nQuants(msnsetSubsetTop3,
                       groupBy=fData(msnsetSubset)$ProteinAccession)
msnsetSubsetTop3 <- combineFeatures(msnsetSubsetTop3,
                                    groupBy=fData(msnsetSubset)$ProteinAccession,
                                    fun="sum", na.rm=TRUE)

exprs(msnsetSubsetTop3) <- exprs(msnsetSubsetTop3) * (3/nSubsetPeps)
# Error in exprs(msnsetSubsetTop3) * (3/nSubsetPeps) :
#  non-conformable arrays

This PR removes utils.colSd, utils.applyColumnwiseByGroup, rewrites nQuants, featureCV, adds rowmean, rowsd (both similar to rowsum) and adds unit tests for all of them.

As side-effect it changes the output of nQuants and featureCV. Now they return a matrix with nrow(x) == sum(levels(group) %in% group) instead of nrow(x) == nlevels(group). I don't think anybody would use the former output format. In the current phase of the release cycle it should be safe to change the dimension of the return values.

Same example as above with the new implementation:

library("MSnbase")
data(msnset)

nQuants(msnset, fData(msnset)$ProteinAccession)
#         iTRAQ4.114 iTRAQ4.115 iTRAQ4.116 iTRAQ4.117
# BSA              3          3          3          3
# ECA0172          1          1          1          1
# ECA0435          2          2          2          2
# ...
# ECA4514          6          6          6          6
# ENO              4          4          3          4

msnsetSubset <- msnset[1:5]
nQuants(msnsetSubset, fData(msnsetSubset)$ProteinAccession)
#         iTRAQ4.114 iTRAQ4.115 iTRAQ4.116 iTRAQ4.117
# BSA              1          1          1          1
# ECA1364          1          1          1          1
# ECA1422          1          1          1          1
# ECA3882          1          1          1          1
# ECA4030          1          1          1          1

msnsetSubsetTop3 <- topN(msnsetSubset,
                         groupBy=fData(msnsetSubset)$ProteinAccession,
                         n=3)
nSubsetPeps <- nQuants(msnsetSubsetTop3,
                       groupBy=fData(msnsetSubset)$ProteinAccession)
msnsetSubsetTop3 <- combineFeatures(msnsetSubsetTop3,
                                    groupBy=fData(msnsetSubset)$ProteinAccession,
                                    fun="sum", na.rm=TRUE)

exprs(msnsetSubsetTop3) <- exprs(msnsetSubsetTop3) * (3/nSubsetPeps)
msnsetSubsetTop3
# MSnSet (storageMode: lockedEnvironment)
# assayData: 5 features, 4 samples
#   element names: exprs
# protocolData: none
# phenoData: none
# featureData
#   featureNames: BSA ECA1364 ... ECA4030 (5 total)
#   fvarLabels: spectrum ProteinAccession ... CV.iTRAQ4.117 (19 total)
#   fvarMetadata: labelDescription
# experimentData: use 'experimentData(object)'
# Annotation:
# - - - Processing information - - -
# Data loaded: Wed May 11 18:54:39 2011
# iTRAQ4 quantification by trapezoidation: Wed Apr  1 21:41:53 2015
# Subset [55,4][5,4] Fri Apr 28 18:13:18 2017
# Selected top 3 features [Fri Apr 28 18:13:57 2017]
# Subset [5,4][5,4] Fri Apr 28 18:13:57 2017
# Combined 5 features into 5 using sum: Fri Apr 28 18:13:57 2017
#  MSnbase version: 2.3.1