This is a test version of Biostars. For the public version, visit https://www.biostars.org.
PFAM double entry for the same PFAM accession with overlapping amino acids, how to keep only one?

Hello,

I am inspecting the in Pfam-A.regions.uniprot.tsv file

for some entries I get double overlapping entry for same accession, example below:

P06753  2       99EAD24C45460A14        8675e9f4c97a6b94eb04b83b56c4a67e        PF00261 68      285     68      284
P06753  2       99EAD24C45460A14        8675e9f4c97a6b94eb04b83b56c4a67e        PF00261 2       84      2       83

However, when I check the accession in InterPro, I get only one entry, from aa 68 to 284.

I cannot find how they choose one entry over the other, in the documentation they say we should use the E-value, but I did not find it in the tsv file, neither in the Pfam-A.clans.tsv.gz. Do you know how I could get the same result as the InterPro UI or if I should keep both entries? Thanks a lot!

interpro domains pfam functional

1 answer

Pfam-A.regions.uniprot.tsv is the raw region table and deliberately carries no scores, which is why you can't apply the E-value rule from that file. InterPro isn't showing a different dataset, it's showing the same matches after its own overlap resolution.

Simplest fix is to skip reconciling it yourself and pull the resolved matches from the InterPro API, since that's what the UI renders:

https://www.ebi.ac.uk/interpro/api/entry/pfam/protein/uniprot/P06753/

If you do want to do it locally you need bitscores, so run hmmscan against Pfam-A.hmm with --cut_ga and keep the higher-scoring hit where two overlap.

Worth noting your two rows barely overlap: 2-84 and 68-285 share only about 17 residues. P06753 is tropomyosin, a long coiled-coil, so that's PF00261 tiling along the sequence rather than two distinct domains, which is the case InterPro collapses to the better-scoring match.

Log in to answer this question.