Dealing with peptides/proteins that aren't detected in all conditions...
nickschurch opened this issue · 2 comments
Proteus really needs to have a good way of dealing with proteins and/or peptides that are not detected in one or more of the conditions. Currently, Proteus includes these data int he data import by limmaDE will not report a fold-change or a significance for them, nor will they be reported in the list of SDE molecules, and not will they be present on the volcano/fold-change-intensity plots that result. For some experiments (notably those involving an IP pulldown) these results are critical tot he experiment and are, arguably, the strongest evidence in the data.
I suggest computing a boolean table describing whether each peptide or protein is detected in each condition and adding this to both the peptide and protein tables (computed during makePeptideTable & makeProteinTable). This can then be easily use to index peptides/proteins detected/undetected in any combination of conditions (a/several functions for this would be useful too). In conjunction with the limmaDE results, these proteins/peptides can also be reported in the SDE list and cal also be added to the plots as points pegged to one/more axes and labelled with a different colour.
Partially done. Added detect field to protein/peptide object with a logical table indicating detection in each condition. Updated the vignette as well. Will deal with plots later.
From version 0.1.8 plotFID takes data present in only one into account. They are now shown (only in binhex=FALSE mode) as orange points at the top and bottom of the plot. This is not possible in plotVolcano as these data lack both p-value and fold change.