The trick with running data through a whole bunch of different engines is generally the pile of peptides where the engines disagree about whether A) there is an identification or (worse) B) that spectrum is matched by different engines to different sequences.
It's similar to the workflow I use for TMT single cells that I called "The Trifecta" that I'm sad to say I can't get to work anymore. At some point some update caused MSFragger and PD to never play together again, and no amount of backtracking seemed to be able to get all the connections re-established for me.
But the reason Trifecta works is circled. All the PSM IDs end up going into Percolator with the same 30-ish score critieria Percolator looks at in each PSM given (probably) the same weights. Then the same thing happens at the peptide group FDR and the protein group FDR and matches end up going to IMP-prtmRS for scoring of the PTMs.
As the developer of 3 of the nodes on the screenshot above (and 2 in the protein level workflow that are employed in Trifecta) it's a safe assumption that this was pretty well throught through and
Those equivalent steps are here. What's cool about TripleToolWF is that you can't get much more different from a search engine tool set than PEAKS, MSFragger, Inferys. They're all looking at those spectra in very different ways. And compiling those results intelligently (not just taking every ID at face value to get a big pile of IDs) seems to help a lot here!
No comments:
Post a Comment