Currently Skyline only runs Percolator at the single-file level which will cause inflated FDR as the number of files increases, and has been shown to lead to libraries for DIA that do not do as well as ones where at least all-files-q-values are used at the peptide level. Mike has also said that Bill Noble no longer feels one should ever use PSM q values for an overall dataset cut-off as we are currently doing.

We need to figure out how to run Percolator on the full set of files at the end of our processing and build the spectral library using the all-files peptide q values, as we do with pdResult files after we found the same problem there.