This seems like a strange problem. Are you describing model estimates or Decoy counts here? I would recommend carefully looking at the model to decoys FDR plots and the individual models both from PeptideProphet for each of the searches and iProphet. It could be a misbehaving search engine or incorrect assumption in the iProphet modeling that would require turning off. I am not sure. If you can share the data I could explore further.
-David On Mar 6, 2015 5:17 AM, "Florian" <[email protected]> wrote: > Dear list, > > I am trying to incorporate multiple database search engines and iProphet > into my data analysis. I run the TPP from commandline with following > commands: > > 1) For single search engines (here: X!Tandem): > xinteract -Ninteract_xtandem.pep.xml -p0.05 -l7 -OARP -dSHUFFLED > -X-m0.02-nK,6.031817-nn,6.031817-H *.pep.xml > ProteinProphet interact_xtandem.pep.xml interact_xtandem.prot.xml XPRESS > EXCLUDE_ZEROS > > 2) To use iProphet on this single database search engines: > xinteract -Ninteract_xtandem.pep.xml -p0.05 -l7 -OARP -dSHUFFLED > -X-m0.02-nK,6.031817-nn,6.031817-H *.pep.xml > c:\Inetpub\tpp-bin\InterProphetParser \Xtandem\interact_xtandem.pep.xml > interact.ipro.pep.xml > ProteinProphet interact.ipro.pep.xml interact_xtandem.ipro.prot.xml XPRESS > EXCLUDE_ZEROS IPROPHET > > 3) For multiple search engines (here: X!Tandem and MSGF+): > c:\Inetpub\tpp-bin\InterProphetParser \Xtandem\interact_xtandem.pep.xml > \MSGF\interact_msgf.pep.xml interact.ipro.pep.xml > RefreshParser interact.ipro.pep.xml %dbase% > ProteinProphet interact.ipro.pep.xml interact_ipro.prot.xml XPRESS > EXCLUDE_ZEROS IPROPHET > > 4) For testing, I also ran ProteinProphet on the combined iProphet output > without using the IPROPHET option: > c:\Inetpub\tpp-bin\InterProphetParser \Xtandem\interact_xtandem.pep.xml > \MSGF\interact_msgf.pep.xml interact.ipro.pep.xml > RefreshParser interact.ipro.pep.xml %dbase% > ProteinProphet interact.ipro.pep.xml interact_ipro.prot.xml XPRESS > EXCLUDE_ZEROS > > In an example analysis, I identified the following number of Proteins > after using <1% FDR based on the ProteinProphet estimation: > 1a) X!Tandem only: 3,657 Proteins (589 single hits) > 1b) MSGF+ only: 3,791 Proteins (614 single hits) > 2a) X!Tandem with iProphet: 3,325 (735 single hits) > 2b) MSGF+ with iProphet: 3,473 Proteins (773 single hits) > 3) X!Tandem, MSGF+ and iProphet: 3,803 Proteins (964 single hits) > 4) X!Tandem, MSGF+ without IPROPHET option: 4,220 proteins (732 single > hits) > > It seems like iProphet reduces the number of protein IDs, while increasing > the single hits. So, if I substract the single hits, I even get less > protein IDs when using two database search engines instead of one. It even > gets worse when I use Myrimatch together with X!Tandem and MSGF+. How can > that be? Did I do some obvious mistake here? Or do I have to switch off > single models of the iProphet? > > If not, is it acceptable to use iProphet output without the IPROPHET > option in ProteinProphet? This way, I can use multiple search engines and > INCREASE my IDs, instead of loosing some. > > Many thanks! > > > > > -- > 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/d/optout. > -- 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/d/optout.
