Peptides Mapped to Wrong proteins

[Skourtis]

The Peptides which are reported in the ion_label_quant.tsv, in the modified sequence, are mapped directly to fasta file based on sequence matching, but often that peptide couldn't have arisen from that protein (because there is no cleavage site e.g. R|K for trypsin before the peptide) in the specfic protein.

E.g. This peptide AELLAGR maps to all these isoforms in the ion_label_quant (and probably downstream) but as you can see from their fasta sequence, They all have a W which is non-tryptic before the sequence. In this case I suspect it arises from a rev_ sequence in the fasta file and then accidentally matches to real proteins.

This run was done with enzymatic trypsin cleavage, (not for example with semi_n_term cleavage), so the R and K before should be respected.

This leads to peptides assigned to more proteins than they should, changing the Razor rule.

[Skourtis]

This also happens with post-translational modifications where a sequence n[42.0106]NGTSM[15.9949]ISLIIPPK, could only possibly arise from a protein which starts with this sequence (since only the [^ was allowed the n[42.0106] mod).
Again you can see that instead of only mapping to P62495_nterm_5204, because it is the n-term of the protein, and the other sequences don't have an R|K before the sequence, it maps to many of them.

[fcyu]

Thanks for your report. Are you using FragPipe 18.0 with Percolator enabled?

Thanks,

Fengchao

[Skourtis]

I'm using Fragpipe v17.1 and for PSM I'm using PeptideProphet for Closed Search.

[fcyu]

Then, the peptide-protein mapping is from PeptideProphet. I am not sure if it is something that we can change/fix.

Best,

Fengchao

[prvst]

Can you send me your .pepXML and pep.xml files?

[prvst]

Thanks

[anesvi]

Refresh parser tool of PeptideProphet maps based on the sequence. I do not think it consider cleavage rules.

In MSFragger 3.5 we remap peptides ourselves so if used with percolator then you have get more accurate mapping, Fengchao?

Get Outlook for iOShttps://aka.ms/o0ukef

---

From: Savvas Kourtis @.> Sent: Tuesday, June 14, 2022 6:42:23 AM To: Nesvilab/FragPipe @.> Cc: Subscribed @.***> Subject: Re: [Nesvilab/FragPipe] Peptides Mapped to Wrong proteins (Issue #718)

External Email - Use Caution

This also happens with post-translational modifications where a sequence n[42.0106]NGTSM[15.9949]ISLIIPPK, could only possibly arise from a protein which starts with this sequence (since only the [^ was allowed the n[42.0106] mod). Again you can see that instead of only mapping to P62495_nterm_5204, because it is the n-term of the protein, and the other sequences don't have an R|K before the sequence, it maps to many of them. [image]https://user-images.githubusercontent.com/51754041/173559134-44fb8478-0981-4a24-8c4b-51bfdd7f1922.png

[image]https://user-images.githubusercontent.com/51754041/173558863-7d12dc38-dd19-4651-8580-19831d6e0ab7.png [image]https://user-images.githubusercontent.com/51754041/173559016-fb352dcd-3361-4823-a047-7f293b52246f.png

— Reply to this email directly, view it on GitHubhttps://github.com/Nesvilab/FragPipe/issues/718#issuecomment-1155016350, or unsubscribehttps://github.com/notifications/unsubscribe-auth/AIIMM63KMKDU3XXFGU342BLVPBOY7ANCNFSM5YXF4V3Q. You are receiving this because you are subscribed to this thread.Message ID: @.***>

---

Electronic Mail is not secure, may not be read every day, and should not be used for urgent or sensitive issues

[prvst]

From your first example, we can see that fragger mapped the peptide AELLAGR to the decoy protein Q15772:

<search_result>
<search_hit peptide="AELLAGR" massdiff="0.00103759765625" calc_neutral_pep_mass="738.4263" peptide_next_aa="C" num_missed_cleavages="0" num_tol_term="2" protein_descr="Striated muscle preferentially expressed protein kinase OS=Homo sapiens OX=9606 GN=SPEG PE=1 SV=4" num_tot_proteins="1" tot_num_ions="12" hit_rank="1" num_matched_ions="6" protein="rev_sp|Q15772|SPEG_HUMAN" peptide_prev_aa="R" is_rejected="0">
<modification_info modified_peptide="AELLAGR[166]">
<mod_aminoacid_mass mass="166.1094" position="7"/>
</modification_info>
<search_score name="hyperscore" value="13.407"/>
<search_score name="nextscore" value="11.721"/>
<search_score name="expect" value="1.057631e+00"/>
</search_hit>
</search_result>

PeptideProphet scored the mapping and added the real protein O15360 as an alternative to the assignement:

<spectrum_query start_scan="44773" uncalibrated_precursor_neutral_mass="738.42645" assumed_charge="2" spectrum="2020LD002_MAGU_002_03_50pto.44773.44773.2" end_scan="44773" index="7009" precursor_neutral_mass="738.4273" retention_time_sec="1375.9601211547852">
<search_result>
<search_hit peptide="AELLAGR" massdiff="0.00103759765625" calc_neutral_pep_mass="738.4263" peptide_next_aa="C" num_missed_cleavages="0" num_tol_term="2" protein_descr="Striated muscle preferentially expressed protein kinase OS=Homo sapiens OX=9606 GN=SPEG PE=1 SV=4" num_tot_proteins="2" tot_num_ions="12" hit_rank="1" num_matched_ions="6" protein="rev_sp|Q15772|SPEG_HUMAN" peptide_prev_aa="R" is_rejected="0">
<alternative_protein protein="sp|O15360|FANCA_HUMAN" protein_descr="Fanconi anemia group A protein OS=Homo sapiens OX=9606 GN=FANCA PE=1 SV=2" num_tol_term="1" peptide_prev_aa="W" peptide_next_aa="V"/>
<modification_info modified_peptide="AELLAGR[166]">
<mod_aminoacid_mass mass="166.1094" position="7"/>
</modification_info>
<search_score name="hyperscore" value="13.407"/>
<search_score name="nextscore" value="11.721"/>
<search_score name="expect" value="1.057631e+00"/>
<analysis_result analysis="peptideprophet">
<peptideprophet_result probability="0.4301" all_ntt_prob="(0.0000,0.0000,0.4301)">
<search_score_summary>
<parameter name="fval" value="0.9580"/>
<parameter name="ntt" value="2"/>
<parameter name="nmc" value="0"/>
<parameter name="massd" value="1.405"/>
<parameter name="isomassd" value="0"/>
</search_score_summary>
</peptideprophet_result>
</analysis_result>
</search_hit>
</search_result>
</spectrum_query>

The switch happened during the filtering step, where we perform something that we call 'protein promotion'. If a PSM maps to a decoy, and has a target entry as an alternative, we flip them, and the assignment becomes a "target" one. That is why you see O15360 in the report.

[fcyu]

We also don't consider the cleavage rules in MSFragger 3.5. I think we should add it in the next version.

Best,

Fengchao

Refresh parser tool of PeptideProphet maps based on the sequence. I do not think it consider cleavage rules. In MSFragger 3.5 we remap peptides ourselves so if used with percolator then you have get more accurate mapping, Fengchao?

[Skourtis]

Hi Everyone!

thank you for your quick and clear responses! I've implemented a peptide-> protein remapping in my scripts while I wait for the next version! Some of the peptides will not map to any real proteins (which will mean that they were actually decoy) but from what I have seen this is extremely rare (10 peptide evidence out of 20000) so shouldn't have an impact on FDR.

Thanks!

[fcyu]

Fixed. Will be available in the next release.

Best,

Fengchao