Triqler#
Triqler is a probabilistic graphical model that propagates error information through all steps from MS1 feature to protein level, employing distributions in favor of point estimates, most notably for missing value imputation [THE2018]. The model outputs posterior probabilities for fold changes between treatment groups, highlighting uncertainty rather than hiding it.
quantms & triqler#
quantms exports the triqler input after the quantification steps in the LFQ analysis (Label-free quantification with DDA (LFQ)). The searchScore is computed as 1-PEP. PEP is the Posterior Error Probability. For DIA analysis, the searchScore is computed as 1-Q.Value. The following table is an example how the exported file should looks like:
run |
condition |
charge |
searchScore |
intensity |
peptide |
proteins |
---|---|---|---|---|---|---|
6 |
heart |
2 |
0.9840915 |
3.275759e07 |
AAAFEQLQK |
O94826 |
Note
The triqler output is stored in the proteomicslfq folder for label-free (Label-free quantification with DDA (LFQ)) experiments and in diannconvert folder for DIA analysis (Data-independent acquisition (DIA) quantification). Currently Triqler does not support labeled experiments. The triqler output generation automatically activates decoy quantification which makes the pipeline a slower.
Some remarks:
For Triqler to work, it also needs decoy PSMs, preferably resulting from a search engine search with a reversed protein sequence database concatenated to the target database. quantms exports the decoy and target proteins into the triqler output.
The intensities should not be log transformed, Triqler will do this transformation for you.
The search engine scores should be such that higher scores indicate a higher confidence in the PSM. quantms uses a transformation of the Posterior error probability (PEP) as 1-PEP for each PSM.
Multiple proteins can be specified at the end of the line, separated by tabs. However, it should be noted that Triqler currently discards shared peptides.
Running Triqler#
Triqler can be run in the quantms output by using the following command:
python -m triqler --fold_change_eval 0.8 out_triqler.tsv
References#
The M, Käll L. Integrated Identification and Quantification Error Probabilities for Shotgun Proteomics. Mol Cell Proteomics. 2019 Mar;18(3):561-570. doi: 10.1074/mcp.RA118.001018. Epub 2018 Nov 27. PMID: 30482846; PMCID: PMC6398204.