Bug in mod reporting if multiple PSMs are tied for rank 1 for a spectrum

[mriffle]

I got this odd result in psm.tsv when using latest philosopher and comet:

QEP2_2017_1003_AZ_135_az583_AZ.58073.58073.2 interact-QEP2_2017_1003_AZ_135_az583_AZ.pep.xml MTAREVMDIDEAREKYAR M[147]TAREVMDIDEAREK[283]YAR 18 2 7061.3000 2753.1887 2753.1887 1377.6016 1377.6016 2354.1358 1178.0752 399.0511 0.4490 0.0010 0.0000 0.0000 1.0000 0.73400000000000 0.0000 0.0000 0.0511 2 3 0.0000 0.0000 0.0000 15K(155.0946), 19M(15.9949), 1M(15.9949), 7M(15.9949), N-term(155.0946) true gi|16128177|ref|NP_414726.1| gi|16128177|ref|NP_414726.1| DNA polymerase III alpha subunit [Escherichia coli str. K-12 substr. MG1655]

What's odd about it are the reported mods:

15K(155.0946), 19M(15.9949), 1M(15.9949), 7M(15.9949), N-term(155.0946)

This is the peptide in question: MTAREVMDIDEAREKYAR

This is odd for a couple reasons:

1. The reported peptide only has two variable mods in it (M[147]TAREVMDIDEAREK[283]YAR), yet psm.tsv is reporting 5 variable mods on this peptide (none of them are static mods)
2. The peptide doesn't have 19 residues, so the reported mod of 19M(15.9949) makes no sense.

Digging into the pepXML generated by commet, I see the following for that spectrum:

 <spectrum_query spectrum="QEP2_2017_1003_AZ_135_az583_AZ.58073.58073.2" spectrumNativeID="controllerType=0 controllerNumber=1 scan=58073" start_scan="58073" end_scan="58073" precursor_neutral_mass="2753.188721" assumed_charge="2" index="51618" retention_time_sec="7061.3">
  <search_result>
   <search_hit hit_rank="1" peptide="GGGNTIEYFVNTTFNYPTMAEAYR" peptide_prev_aa="K" peptide_next_aa="V" protein="gi|90111670|ref|NP_418397.2|" num_tot_proteins="1" num_matched_ions="2" tot_num_ions="46" calc_neutral_pep_mass="2886.317017" massdiff="-133.128296" num_tol_term="2" num_missed_cleavages="0" num_matched_peptides="245742">
    <modification_info modified_peptide="n[156]GGGNTIEYFVNTTFNYPTM[147]AEAYR" mod_nterm_mass="156.102425">
     <mod_aminoacid_mass position="19" mass="147.035385" variable="15.994900" source="param"/>
    </modification_info>
    <search_score name="xcorr" value="0.449"/>
    <search_score name="deltacn" value="0.000"/>
    <search_score name="deltacnstar" value="0.000"/>
    <search_score name="spscore" value="5.0"/>
    <search_score name="sprank" value="2"/>
    <search_score name="expect" value="7.34E-01"/>
   </search_hit>
   <search_hit hit_rank="1" peptide="MTAREVMDIDEAREKYAR" peptide_prev_aa="K" peptide_next_aa="G" protein="gi|16128177|ref|NP_414726.1|" num_tot_proteins="1" num_matched_ions="2" tot_num_ions="34" calc_neutral_pep_mass="2354.135822" massdiff="399.052899" num_tol_term="2" num_missed_cleavages="3" num_matched_peptides="245742">
    <modification_info modified_peptide="MTAREVM[147]DIDEAREK[283]YAR">
     <mod_aminoacid_mass position="7" mass="147.035385" variable="15.994900" source="param"/>
     <mod_aminoacid_mass position="15" mass="283.189563" variable="155.094600" source="param"/>
    </modification_info>
    <search_score name="xcorr" value="0.449"/>
    <search_score name="deltacn" value="0.001"/>
    <search_score name="deltacnstar" value="0.001"/>
    <search_score name="spscore" value="6.7"/>
    <search_score name="sprank" value="1"/>
    <search_score name="expect" value="7.34E-01"/>
   </search_hit>
   <search_hit hit_rank="1" peptide="MTAREVMDIDEAREKYAR" peptide_prev_aa="K" peptide_next_aa="G" protein="gi|16128177|ref|NP_414726.1|" num_tot_proteins="1" num_matched_ions="2" tot_num_ions="34" calc_neutral_pep_mass="2354.135822" massdiff="399.052899" num_tol_term="2" num_missed_cleavages="3" num_matched_peptides="245742">
    <modification_info modified_peptide="M[147]TAREVMDIDEAREK[283]YAR">
     <mod_aminoacid_mass position="1" mass="147.035385" variable="15.994900" source="param"/>
     <mod_aminoacid_mass position="15" mass="283.189563" variable="155.094600" source="param"/>
    </modification_info>
    <search_score name="xcorr" value="0.449"/>
    <search_score name="deltacn" value="0.001"/>
    <search_score name="deltacnstar" value="0.000"/>
    <search_score name="spscore" value="6.7"/>
    <search_score name="sprank" value="1"/>
    <search_score name="expect" value="7.34E-01"/>
   </search_hit>
   <search_hit hit_rank="2" peptide="GGGNTIEYFVNTTFNYPTMAEAYR" peptide_prev_aa="K" peptide_next_aa="V" protein="gi|90111670|ref|NP_418397.2|" num_tot_proteins="1" num_matched_ions="2" tot_num_ions="46" calc_neutral_pep_mass="2731.222417" massdiff="21.966304" num_tol_term="2" num_missed_cleavages="0" num_matched_peptides="245742">
    <modification_info modified_peptide="GGGNTIEYFVNTTFNYPTM[147]AEAYR">
     <mod_aminoacid_mass position="19" mass="147.035385" variable="15.994900" source="param"/>
    </modification_info>
    <search_score name="xcorr" value="0.449"/>
    <search_score name="deltacn" value="0.100"/>
    <search_score name="deltacnstar" value="0.000"/>
    <search_score name="spscore" value="5.0"/>
    <search_score name="sprank" value="2"/>
    <search_score name="expect" value="7.46E-01"/>
   </search_hit>
   <search_hit hit_rank="3" peptide="HPPTDYVEEGHDSFYVLFGNPNAAKFDK" peptide_prev_aa="K" peptide_next_aa="T" protein="rev_gi|16131793|ref|NP_418390.1|" num_tot_proteins="1" num_matched_ions="2" tot_num_ions="54" calc_neutral_pep_mass="3193.478129" massdiff="-440.289408" num_tol_term="2" num_missed_cleavages="1" num_matched_peptides="245742">
    <search_score name="xcorr" value="0.404"/>
    <search_score name="deltacn" value="1.000"/>
    <search_score name="deltacnstar" value="0.000"/>
    <search_score name="spscore" value="3.6"/>
    <search_score name="sprank" value="4"/>
    <search_score name="expect" value="2.52E+00"/>
   </search_hit>
  </search_result>
 </spectrum_query>

So, what it looks like to me, is that philosopher is grabbing all the reported modifications for all the rank 1 PSMs for this spectrum, and listing them all one the same line as a single psm in psm.tsv. It does this even if the peptides are not the same--which is extra bad.

IMO, the preferred behavior would be to list each of the rank 1 hit as separate lines in psm.tsv.

[mriffle]

Actually, just remembered PeptideProphet only puts a probability on one of the rank 1 PSMs, so I think reporting all of them wouldn't work...

[prvst]

Hi @mriffle. Thanks for reporting this. The idea was never to collapse the top hits ties from Comet, I developed our logic around MSFragger which does not produce results with ties. I'll work on a fix for this.