ProteinInference

A peptide identified by PSMStatistics often fits more than one protein, for example two splice variants of the same gene. ProteinInference maps the identified peptides back to proteins and reports protein groups: sets of proteins that share their peptide evidence. Each group gets a score from its peptides and a decoy score from the reversed proteins, and from these a protein level q-value.

Two parameters decide how groups are formed and which peptides count for them. Take two proteins pA and pB, each found by one unique peptide and both by a third shared peptide. Protein (the integration strictness) decides whether pA, pB and the group pA;pB are all reported or whether overlapping groups are merged into the smallest set that explains every peptide. Peptide (the peptide usage) decides whether the abundance of pA;pB later comes from the shared peptide only or also from the peptides unique to pA and pB. The figure shows the outcomes for every combination.

Inputs and outputs

The GFF3 file tells the tool which proteins are transcripts of the same gene. With it the Class column (the peptide evidence class) separates groups made of isoforms of one gene from groups that mix unrelated proteins. Without -g every database protein counts as its own gene and the class carries no such information.

For every input run.qpsm the tool writes run.prot, tab separated with a header and the columns ProteinGroup (protein identifiers joined by ;), PeptideSequence (the peptides of the group found in this run, joined by ;), Class, TargetScore, DecoyScore and QValue. JoinQuantPepIonsWithProteins and AddDeducedPeptides read .prot files.

With -dc the tool saves a chart of scores against q-values, as QValueGraph.html in the output directory when the files are grouped and as run.prot_QValueGraph.html per run otherwise. Logs go to ProteinInference_log.txt and a few ProteinInference_<step>_log.txt files in the output directory. There is no parallelism flag.

Parameters

The default file is ProteinInferenceParams.json.

Writing a parameter file

open ProteomIQon
open ProteomIQon.Domain
open BioFSharp.Mz

let proteinInferenceParams : Dto.ProteinInferenceParams =
    {
        ProteinIdentifierRegex = @"Cre\S+"
        Protein                = ProteinInference.IntegrationStrictness.Maximal
        Peptide                = ProteinInference.PeptideUsageForQuantification.Minimal
        GroupFiles             = true
        GetQValue              = QValueMethod.LogisticRegression FDRMethod.MAYU
    }

// Replace the temp folder with your project folder.
let outputPath = System.IO.Path.Combine(System.IO.Path.GetTempPath(), "ProteinInferenceParams.json")

Json.serializeAndWrite outputPath proteinInferenceParams

Running the tool

Install with dotnet tool install --global ProteomIQon.ProteinInference, then infer the proteins of one run:

proteomiqon-proteininference -i path/to/run.qpsm -d path/to/database.db -o path/to/output -p path/to/ProteinInferenceParams.json

Several runs grouped together, with the annotation for the evidence class:

proteomiqon-proteininference -i path/to/run1.qpsm path/to/run2.qpsm path/to/run3.qpsm -d path/to/database.db -g path/to/proteome.gff3 -o path/to/output -p path/to/ProteinInferenceParams.json

A whole directory, with the q-value chart:

proteomiqon-proteininference -i path/to/qpsmFolder -d path/to/database.db -o path/to/output -p path/to/ProteinInferenceParams.json -dc

All flags:

proteomiqon-proteininference --help