No decoy PSMs were detected

[wfondrie]

I received the following question as a Twitter DM:

Hi Will,
I am trying (to try out) mokapot and get this error:

[INFO] mokapot version 0.8.3
[INFO] Written by William E. Fondrie (wfondrie@uw.edu) in the
[INFO] Department of Genome Sciences at the University of Washington.
[INFO] Command issued:
[INFO] /Users/pwilmart/opt/anaconda3/envs/py3.9/bin/mokapot --min_length 7 DRUK-951_TMT1_Phospho_CML_TMT17plex_1D-LC_0pt57ug_PSRtrapPSRcol_140min_20220511_OT.pin
[INFO] 
[INFO] Starting Analysis
[INFO] =================
[INFO] Parsing PSMs...
[INFO] Reading DRUK-951_TMT1_Phospho_CML_TMT17plex_1D-LC_0pt57ug_PSRtrapPSRcol_140min_20220511_OT.pin...
[INFO] Using 21 features:
[INFO]   (1)lnrSp
[INFO]   (2)deltLCn
[INFO]   (3)deltCn
[INFO]   (4)lnExpect
[INFO]   (5)Xcorr
[INFO]   (6)Sp
[INFO]   (7)IonFrac
[INFO]   (8)Mass
[INFO]   (9)PepLen
[INFO]   (10)Charge1
[INFO]   (11)Charge2
[INFO]   (12)Charge3
[INFO]   (13)Charge4
[INFO]   (14)Charge5
[INFO]   (15)Charge6
[INFO]   (16)enzN
[INFO]   (17)enzC
[INFO]   (18)enzInt
[INFO]   (19)lnNumSP
[INFO]   (20)dM
[INFO]   (21)absdM
[INFO] Found 259032 PSMs.
[INFO]   - 259032 target PSMs and 0 decoy PSMs detected.
Traceback (most recent call last):
  File "/Users/pwilmart/opt/anaconda3/envs/py3.9/bin/mokapot", line 10, in <module>
    sys.exit(main())
  File "/Users/pwilmart/opt/anaconda3/envs/py3.9/lib/python3.9/site-packages/mokapot/mokapot.py", line 59, in main
    datasets = parse(config.psm_files)
  File "/Users/pwilmart/opt/anaconda3/envs/py3.9/lib/python3.9/site-packages/mokapot/parsers/pin.py", line 158, in read_pin
    return LinearPsmDataset(
  File "/Users/pwilmart/opt/anaconda3/envs/py3.9/lib/python3.9/site-packages/mokapot/dataset.py", line 490, in __init__
    raise ValueError("No decoy PSMs were detected.")
ValueError: No decoy PSMs were detected.

I get the same error on my Mac OS or in Windows 10. I made the FASTA with concatenated reversed decoys using the function in mokapot. I see accessions that start with “decoy_” in the pin file. I am using a clean Python 3.9 environment in Anaconda. Any idea what might be up? Since this is my first try, I did not want to post on the discussion site. It might be something dumb that I am doing.

I noticed that the command line use allows specifying a decoy prefix (I usually have “REV_” in my FASTA files) but did not see an option to set that in the Python API.

[wfondrie]

It looks like all values of the "Label" column in the pin file are 1, indicating all of the PSMs are targets. This is the column that is used by mokapot (and Percolator) to distinguish between target and decoy PSMs. Please indicate decoy PSMs with a -1 and it should work.

As for the decoy prefix, this is only used for protein-level FDR estimates. To do the "picked-protein approach", mokapot needs to be able to pair target proteins to their decoy counterparts in the FASTA file, which is where this column comes into play - it has no effect on PSM or peptide level FDR estimates.

To specify a different decoy prefix in the Python API, use the decoy_prefix argument in http://mokapot.read_fasta(): https://mokapot.readthedocs.io/en/latest/api/functions.html#mokapot.read_fasta

[wfondrie]

Followup question from Twitter:

I would have never guessed that I have to set the target or decoy status in the PIN file when the decoy prefix of the accessions is a settable parameter. That seems like a strange burden to put on the user. Is this described in the docs? I might have skimmed over it. How are the decoys handled if a pep xml file is read in?

[wfondrie]

That seems like a strange burden to put on the user. Is this described in the docs?

Alas, it may be necessary when users are not using data that originates from a protein FASTA file---for instance from a spectral library search, or a database of peptides. The necessary columns for files in the "Percolator input format" are describe in the mokapot docs here, and in the Percolator docs here.

I might have skimmed over it. How are the decoys handled if a pep xml file is read in?

When reading the PepXML file, the decoy_prefix is indeed used to detect decoy PSMs.