search

UCLouvain-CBIO / scp

Single cell proteomics data processing
https://uclouvain-cbio.github.io/scp/index.html
21 stars 2 forks source link

Error in computeFDR #4

Closed lgatto closed 4 years ago

lgatto commented 4 years ago

Running example(computeFDR):

> example(computeFDR)

cmpFDR> scp1 <- computeFDR(scp1,
cmpFDR+                    i = 1,
cmpFDR+                    groupCol = "Sequence",
cmpFDR+                    pepCol = "PEP")
Error in computeFDR(scp1, i = 1, groupCol = "Sequence", pepCol = "PEP") : 
  'PEP' must link to a probability in (0, 1)
cvanderaa commented 4 years ago

Oh yes, I wanted to discuss this with you.

I have put a check that makes sure the users supply a probability, but it seems that MaxQuant's PEP computation can go beyond 1. See for example:

range(rowData(scp1[[1]])$PEP)
[1] 1.6819e-21 2.8325e+00

An easy fix is to use the DART-ID corrected PEPs, dart_PEP.

Should I remove the check or change PEP to dart_PEP?

cvanderaa commented 4 years ago

Also you will run in the following warning:

 Features are sorted. This is only needed when replicating the SCoPE2 analysis
See also https://github.com/SlavovLab/SCoPE2/issues/3

This is because part of the implementation is adapted to exactly replicate the FDR computation from the SCoPE2 analysis (see this issue)

lgatto commented 4 years ago