It should be number(decoys)/number(targets) irrespective of concatenated or separate searches.
On Wed, Nov 24, 2010 at 5:20 AM, Amit Yadav <[email protected]> wrote: > Hi, > > From what I understand, the simplest ways to calculate FDR using a > Target-Decoy search are:- > > 1. Run a Concatenate target and decoy database search. The formula > would be : 2 *number(decoys)/ number(target+decoys) - See Elias > and Gygi , Nat Methods. 2007 Mar;4(3):207-14. > > 2. Run Separate target and decoy searches. The formula would be: > number(decoys)/number(targets). See Kall et al J Proteome Res. 2008 > Jan;7(1):29-34 > > I think your formula should be the first one. Members, please correct > me if I am wrong. > > Regards, > > Amit Kumar Yadav > Senior Research Fellow (SRF-CSIR) > IGIB, New Delhi (India) > > MassWiz Web server > MassWiz sourceforge project > MassWiki > > > > On Wed, Nov 24, 2010 at 6:39 PM, Bjorn <[email protected]> wrote: >> >> Hello all, >> >> I am trying to find my way through the different interpretations and >> calculations of the FDR, but I'm a bit lost where it comes to the >> PeptideProphet. >> >> So far, I know that the simplest calculation is (decoy_hits/ >> target_hits). So, I ran X!tandem on my dataset (target DB + >> concatenated decoy DB) and then ran PeptideProphet on the resulting >> pep.xml file (not setting the decoy option!). >> I then set a certain probability threshold (say 0.85) and looked at >> the number of spectra (e.g. 1200 of 1500). Setting the protein text to >> decoy gives me, e.g. 50 hits. The FDR for a prob. of 0.85 would then >> be 50/1150. If I look at the table displayed in the PeptideProphet, >> the FDR I found is almost spot-on with the number in the table. The >> thing is that if I lower my probabilty threshold, the FDRs in the >> table stay relatively low, while the FDRs I calculate with the above >> formula go up. Is there any extra correction being done in the >> PeptideProphet (like correct for the percentage of incorrect target >> hits (PIT) as described by Kall et al in 2008)? >> Or am I missing something vital here? >> >> Furthermore, how is the estimated number of correct assignments >> calculated? I can't seem to find the correct formula to come to that >> number? I assume it is related to the surface under the curve plotted >> with the correct identifications? >> >> Many questions, but I hope someone can help me with this. >> >> Best regards, >> >> bjorn >> >> -- >> You received this message because you are subscribed to the Google Groups >> "spctools-discuss" group. >> To post to this group, send email to [email protected]. >> To unsubscribe from this group, send email to >> [email protected]. >> For more options, visit this group at >> http://groups.google.com/group/spctools-discuss?hl=en. >> > > -- > You received this message because you are subscribed to the Google Groups > "spctools-discuss" group. > To post to this group, send email to [email protected]. > To unsubscribe from this group, send email to > [email protected]. > For more options, visit this group at > http://groups.google.com/group/spctools-discuss?hl=en. > > -- You received this message because you are subscribed to the Google Groups "spctools-discuss" group. To post to this group, send email to [email protected]. To unsubscribe from this group, send email to [email protected]. For more options, visit this group at http://groups.google.com/group/spctools-discuss?hl=en.
