Dear Brendan and Mike: Thanks very much for your replies, I have tried my dataset(phospho and membrane) by peptideprophet and proteinprophet, but after that, so few phophoproteins are shown to be confident at protein FDR<=1% and strangely, most of the peptides in the phophproteins have no phosphorylation. So now i am wondering if my phospho- (or even membrane) protein dataset is fit for proteinprophet?
Recently, i read one article from here: http://www.matrixscience.com/blog/does-protein-fdr-have-any-meaning.html Now my question is that, since some reviewers ask me to provide peptide FDR and protein FDR ( i forgot to write both when submitting my paper), if i just use peptide FDR (by percolator) and get the protein by confident peptide ids and at least one unique peptide, is it proper to describe the confident protein this way instead of providing protein FDR? Many thanks, -Peter On Thursday, November 28, 2013 1:28:51 AM UTC+8, Brendan MacLean wrote: > > Hi Peter, > For more information on what Mike explains, I recommend this paper: > > http://www.ncbi.nlm.nih.gov/pmc/articles/PMC2773710/ > > Also, be careful about distinguishing whether your FDR<=1% is at the > unique peptide level or the peptide spectrum match (PSM) level. The same > phenomenon occurs between PSM FDR and unique peptide FDR as between unique > peptide FDR and protein FDR. Namely, true discoveries are more likely to > be repeated, and therefore 1% FDR at the PSM level will grow above 1% for > unique peptides and again further for proteins, as Mike noted. > > Your choice to include only proteins with >1 hit does improve this growth > in FDR, by applying a secondary filter, but it is by no means foolproof > with modern mass spectrometers and multi-run data sets, depending on the > number of spectra you collected versus the number of proteins considered in > your search. I have seen million spectrum data sets where FDR was > seriously underestimated against a human FASTA with proteins for ~23,000 > genes. > > A good way to gain some insight on your true FDR is to have decoy proteins > in your FASTA that are never exposed to any of your search criteria. This > allows you to double-check FDR estimates from the statistical tools you > use. Once you gain familiarity with your tools, this may be come less > necessary, but I have seen statistical tools get these numbers wrong, for > exactly these reasons. > > Good luck with your FDR statistics. > > --Brendan > > On Tuesday, November 26, 2013 8:52:25 PM UTC-8, zion wrote: >> >> Dear All: >> >> If there are some confident peptide ids at FDR<=1% at the peptide level, >> when inferring proteins by confident peptides and at least one unique >> peptide, CAN i say that at the protein level, its FDR<=1%? >> Many thanks, >> >> -Peter >> > -- You received this message because you are subscribed to the Google Groups "spctools-discuss" group. To unsubscribe from this group and stop receiving emails from it, send an email to [email protected]. To post to this group, send email to [email protected]. Visit this group at http://groups.google.com/group/spctools-discuss. For more options, visit https://groups.google.com/groups/opt_out.
