Precursors identified as HUMAN/ECOLI/HUMAN depending on the module

[mlocardpaulet]

I looked into Spectronaut output of the bulk DIA (Astral) and single cell Astral modules.
I find precursors that are found in both modules but not labelled with the same species. They seem to be ECOLI in the 3-species mix (bulk), and HUMAN in the single cell (no ecoli in the mix). I suspected that it comes from the fact that ECOLI are absent from the fasta in the single cell module, but some are identified as YEAST on one side and HUMAN on the other side...

Here is the list:

AVAVVVDPIQSVK/2, DAGTIAGLNVLR/2, FELSGIPPAPR/2, FLIDGFPR/2, FLVMDEADR/2, FSVGGMTDVAEIK/2, FSVGGM[Oxidation]TDVAEIK/2, FTATASLGVIHK/3, FTQAGSEVSALLGR/2, FVIGGPQGDAGLTGR/2, GSELQNYFTK/2, GVC[Carbamidomethyl]TEAGMYALR/2, GVILYGPPGTGK/2, HEELLSK/2, HITIFSPEGR/2, IGLFGGAGVGK/2, IGPLGLSPK/2, IHTDFEK/2, ILC[Carbamidomethyl]DVPC[Carbamidomethyl]SGDGTMR/2, ILC[Carbamidomethyl]DVPC[Carbamidomethyl]SGDGTM[Oxidation]R/2, LDDDGLIAPGVR/2, LGANSLLDLVVFGR/2, LSDLLDWK/2, MGHAGAIIAGGK/2, NFGLYNER/2, NIGISAHIDSGK/2, NMSVIAHVDHGK/2, NMSVIAHVDHGK/3, TIAM[Oxidation]DGTEGLVR/2, TNC[Carbamidomethyl]LDC[Carbamidomethyl]LDR/2, TTPSYVAFTDTER/2, VEIIANDQGNR/2, WIDFDNDYK/2, YMTDGMLLR/2, YMTDGM[Oxidation]LLR/2, DELTLEGIK/2, ELFVM[Oxidation]AR/2, IINEPTAAAIAYGLDK/3, IINEPTAAAIAYGLDKK/3, NM[Oxidation]SVIAHVDHGK/3, VTILGHVQR/2

What I don't understand is why these are not labelled multispecies? Or is there something that I miss?
I think that this should be explored further.

---

What I did to get there:
I retrieved the data sets from the four following hashes:
"19525e795d412acde41fc81b345488e5b6e39e94"
"2eb6ecaa7ea57027adfdfd5db564f0a3d536f494"
"4d93f6f4811c1771a34c99d1c0162968802a3e5b"
"fa30a3e95f3e6b62199e8660fd0afe570fa1d2bc"
I merged using the precursors+species. I looked for the precursors that I found in several rows (=different species in different hashes).

[mlocardpaulet]

Exploring this a bit further, we noticed that this is not specific to Spectronaut:
I also see it with FragPipe:
single cell module = 47b6a176bde27925e227ab037951e6c675445863
bulk astral = 1cd54fe0ba5d6c98ef47688414c4f4c409a6eecf

Protein inference could be the issue. If the protein groups that are reported are results of the inference algorithm. I mean: if there are no other unique peptides for the yeast protein in one search, then the accession could be kicked out of the group.
An alternative would be to have our own dictionary (peptide sequence <-> species), and use it instead of the accessions in the protein group field of the search output to attribute species. Not sure we want to do this.
In any cases, this has little to no impact of the results.

[rodvrees]

So, I looked into it a little bit more. Some differences are due to the fact that some points have used a different fasta:

• 19525e795d412acde41fc81b345488e5b6e39e94
• fa30a3e95f3e6b62199e8660fd0afe570fa1d2bc
• 59abded256fdd37389d6f2a42adf69aeb1702f1a
• 484fc882d8b74eef29787e1db400d6d0af307d4d

These points should/will be removed.
Others could be due to protein inference.

[rodvrees]

I have removed the points. Proteobench/Results_quant_ion_DIA_lowinput#12