Support for Percolator output files

[wsnoble]

It would be great if Percolator PSM-level tab-delimited output files could be supported by FlashLFQ. I looked into this, and there are two major and several minor challenges associated with making this happen.

The first major issue is that Percolator does not include the mzML file name, but only an integer file index, in the output file. This is obviously problematic, and it's something we can look into fixing on the Percolator end of things.

The second major issue is that Percolator does not include retention time. This is harder for us to fix, because this information is also not included in the outputs of many common search engines. It seems like, if you have the scan numbers in the Percolator output, it should be feasible for FlashLFQ to grab the RT from the mzML file. Is this doable?

The other minor issues are that Percolator does not have a "Base Sequence" column and that Percolator uses comma-delimited protein ID lists, rather than semi-colon delimited lists. These latter ones, along with differences in column naming, should be easy to handle.

Here is a tiny sample Percolator output file.
sample.txt

[trishorts]

Thanks for the request Bill. Students are evaporating inverse to the proximity of the holiday. But we'll engage on this and see how we can accommodate you.

[wsnoble]

Ah, I thought "Scan" and "Retention Time" were two different column IDs, but it turns out it's a single column called "Scan Retention Time." Hence, my suggestion above does not make sense. Is there any way you could have an option to provide scan numbers instead of RTs?

[trishorts]

Sorry for the slow response. Just back from holiday.

I'd love to say that we'd accommodate you but I don't see an easy way to do it. All of the peak finding, etc. is based on time and time windows. With scan number, there is no straightforward way to convert. The time between MS1 scans is variable based on the number of MSn scans between, and that changes throughout the run. So, there is not a good way to look w/in a scan range. And the time for MS1s and MSns also varies, so no simply calculation can be performed. We'd need a time for each kind of scan. The easiest fix for us would be for you to add an additional column of retention time in minutes that percolator otherwise ignores. I'm happy to entertain other options if you can suggest some.

[wsnoble]

I think my suggestion perhaps was not clear. I totally agree that you need to get the RT associated with each scan. My suggestion was to enable a mode in which the input file specifies scan numbers and then the RT is read from the given mzML file.

For example, if you are using pyteomics to parse an mzml file, you can get the RT for each scan as follows:

from pyteomics.mzml import MzML

def get_rts(mzml_filename):
    "Extract retention times from an mzml file."

    rts = {} # Key = scan number, value = retention time
    mzml = MzML(mzml_filename)
    for scan in mzml:
        rts[scan["index"]] = scan["scanList"]["scan"][0]["scan start time"]
    sys.stderr.write(f"Read {len(rts)} RTs from {mzml_filename}.\n")

    return(rts)

Requiring Percolator to report the RT is problematic because Percolator does not typically have access to the original mzML file. So we would have to require that every search engine report RT in its output file format, and that Percolator parse and report these RTs in its output. It seems much easier, since FlashLFQ receives the mzML file as input, for it to parse the RTs directly out of the given mzML files. Does this make sense?

[trishorts]

Yes. I will make an attempt.
What are your plans for including filenames with file path in the percolator output? currently, there is only an integer. would you require users to make the substitution?

[wsnoble]

I am hoping we can add this functionality to Percolator directly. I'm waiting to hear from Lukas about this:

percolator/percolator#322

[trishorts]

I saw that issue cuz I am following the percolator issues. Glad you raised it. I'm already working on the coding updates to flashlfq that will allow retention time recovery from scan number.

can you confirm that "spectrum neutral mass" is the experimental neutral mass
and "peptide mass" is the theoretical neutral mass?

I think I have that right but two minutes on google didn't eliminate my doubts.

[wsnoble]

Yes, that's right.

"The computed mass of the spectrum precursor at the given charge state. This is equal to the precursor m/z minus the mass of a proton (1.00727646677 Da), all multiplied by the charge."

"The mass of the peptide sequence, computed as the sum of the amino acid masses plus the mass of water (18.010564684 Da or 18.0153 Da, depending on whether we are using monoisotopic or average mass)."

http://crux.ms/file-formats/txt-format.html

[trishorts]

can you provide me with one raw/mzml file and its companion output so that I can test my code? I have one remaining concern about whats in the sequence column. we need to be able to parse that and compute peptide mass. The trick will be how ptms are annotated.

[wsnoble]

Here you go! Lemme know if you need more info.
percolator.target.psms.txt

(The raw file is too big to be attached -- will try to get it to you some other way.)

[trishorts]

thanks. i can make a box folder or you can. please invite mrshortreed@wisc.edu

[wsnoble]

Here is a link: https://drive.google.com/file/d/1OTLCgkqcsA2ij8H8edY08-DZG1KPOwxb/view?usp=sharing

[trishorts]

ptms annotated by mass shift sub-optimal.......[15.99] [79.97] we'll see what i can do

[trishorts]

species?

[wsnoble]

Human. It's from here:

http://www.ncbi.nlm.nih.gov/pubmed?term=26951766

ftp://massive.ucsd.edu/MSV000079437/raw/