Follow-up question about PSM-level FDP calculation with the beta version

[lemeng-li]

Hello,

This is a follow-up question from my previous issue. After that discussion, I received a beta version of FDRBench, and I am using that version for this test.

I am now testing the PSM-level FDP calculation using UP000005640_9606.fasta database. The calculation can run, but I have two questions about the input and output.

1. Is the protein column required for PSM-level FDP?

From the documentation, I thought the input should contain:

peptide
mod_peptide
charge
q_value
score

But when I used these columns, I got this error:

Error running FDRBench: Column protein is not present in table ...
java.lang.IllegalStateException: Column protein is not present in table ...

After I added a protein column from the MS-GF+ output, FDRBench could run successfully.

So I am wondering if protein is also a required column for the PSM-level calculation in the beta version. I did not see this column mentioned in the current input format documentation.

2. n_p is always 0

After adding the protein column, the program can finish, but all of the n_p values are 0. Because of this, combined_fdp and lower_bound_fdp are also always 0.

For example:

My input file contains:

peptide
mod_peptide
charge
q_value
score
protein

Then I simplified the MS-GF+ output into the required FDRBench input format.

So I am mainly wondering whether I am missing any required input format or preprocessing step for the PSM-level calculation in the beta version. In particular, I am not sure how FDRBench recognizes the entrapment PSMs for n_p, and whether the protein column needs a specific format.

Since n_t has counts but n_p stays at 0, I am wondering if the entrapment PSMs are not being recognized correctly.

I also attached the code I used and the related input/output files, in case they are helpful for checking the problem.

/Library/Java/JavaVirtualMachines/jdk-24.jdk/Contents/Home/bin/java -jar .../fdrbench-1.2.0-beta/fdrbench-1.2.0-beta.jar -i ".../UP000005640_9606-2_entrapment_pep_input (1).tsv" -o ".../UP000005640_9606-2_entrapment_pep_input (1)_fdp_psm.csv" -level psm -pick first -r 1 -seed 2000

UP000005640_9606-2_entrapment_pep_input (1).tsv

Thank you!

[wenbostar]

Hi, thanks for the detailed report — both issues are explained by the same thing.

1. Is protein required?

Only when you don't pass -pep. Without a peptide list file, FDRBench infers entrapment status from the protein ID, so the column becomes required. With -pep (the sequence/decoy/peptide_type file from database generation), PSM-level runs fine with no protein column at all.

2. Why n_p is 0. Your protein values are raw MS-GF+ output:

sp|LMNGEADAMSLDGGFVYLAGK_target|LMNGEADAMSLDGGFVYLAGK_target(pre=-,post=-)

FDRBench looks for the entrapment label _p_target at the end of the protein ID (default -entrapment_pos 1). The trailing (pre=-,post=-) pushes the label off the end, so no row is recognized as an entrapment hit — hence n_p=0 and zero FDP everywhere.

Two ways to fix it:

• strip (pre=...,post=...) from the protein column, or
• pass -pep , which skips the protein column entirely