How to output multiple PSMs per spectrum? about percolator

[glormph]

I've poked around a bit more, and it only happens using -Y for target/decoy competition (I'm searching using a concatenated DB with multiple files so I use -Y to not use mixmax due to duplicate scan numbers), so I'm wondering if it is intentional. On the whole, of a dataset of 22 PSMs, 13 were retained in my tests. Indeed, there are only 13 scan numbers in the input, but multiple sequences and 22 different psm_ids, which differ in their SIR (SpectrumIdentificationResult id?). When not using -Y all of them are in the XML output.

I guess this kind of explains it: #152 (comment) , but I'm still wondering if one should take into account the entire PSM ID or not.

[MatthewThe]

I think you can get the behavior you want by setting --search-input concatenated. You don't need the -Y flag in that case (it will just be ignored).

[glormph]

Aha, yes that indeed works! It is a bit quieter in the stderr about the FDR method (it says "separate searches input detected, but overridden by -I flag...". I don't know what the difference is between -Y and -I though, does it still do the relevant FDR method for concatenated searches?

[MatthewThe]

Yes, --search-input concatenated uses the FDR method for concatenated searches. We specifically introduced this flag to deal with non-standard scenarios such as yours.

[glormph]

Too many options not to get confused :)
So, if I have understood correctly, I used to do -Y on my concatenated search, which avoids mixmax, but is actually meant for when one runs separated searches and do the competition step inside percolator? So when I instead run -I concatenated it will use the same FDR (non mixmax) but not do actual competition based on PSM scan numbers.
And when one runs msgf2pin without -m, the behaviour of -Y and -I concatenated would be more or less identical, since it does not have multiple solutions for a given scan numbers.

Hope I have understood it now, thanks for the quick help! :)

[MatthewThe]

Yes, that's correct :)

[glormph]

Great, thanks very much!