Pfam not completely working

[jamesmkrieger]

For example, the AMPAR test only gives 2 out of 4 pfam domains at least sometimes

[jamesmkrieger]

@atbogetti, any chance you could check this too please seeing as you’ve just been over the pfam code changes?

[atbogetti]

@jamesmkrieger Yes, I have actually started to look into this already. I will figure it out.

[jamesmkrieger]

Thanks! Glad to have you on board

[atbogetti]

@jamesmkrieger I tried the pfam tests and also encountered failures. I have narrowed down the source of the issue to this line:

ProDy/prody/database/pfam.py

Lines 421 to 422 in 412c8fa

	ags[i] = ag.select('resindex {0} to {1}'.format(
	resiStart, resiEnd))

. For some reason, that way of selecting the residues is causing the following error: TypeError: can't multiply sequence by non-int of type 'Forward'. I tried manually inputting the range and got the same error, though when I just use a single number as the selection (for instance, resindex 10) it works just fine. Do you know if something changed in the residue selection language recently? I can dig deeper but just wanted to see if this may be familiar to you before proceeding.

[jamesmkrieger]

I don’t think it changed. The selection parser is too complicated to change really. Maybe something changed in pyparsing though

[atbogetti]

Bingo. I downgraded pyparsing to v3.1.1 (current is 3.1.2) and it works now. The pfam tests also all pass without any failures.

[jamesmkrieger]

Great!

[jamesmkrieger]

Can you revert #1816 and see if they still pass please?

Also, please can you check if this fixes the problems with the latest numpy versions too (see #1655)?

[atbogetti]

@jamesmkrieger Yes I will check those and get back to you.

[jamesmkrieger]

Bingo. I downgraded pyparsing to v3.1.1 (current is 3.1.2) and it works now. The pfam tests also all pass without any failures.

Actually, that's not true. The pfam tests pass because I changed them. The original tests still fail if we change the pyparsing version (see https://github.com/prody/ProDy/actions/runs/8294288857?pr=1845), so I don't think that's actually the issue here although it's definitely an issue that we have to fix (see also #1844).

[jamesmkrieger]

Can you revert #1816 and see if they still pass please?

This was the wrong number. The one I meant was #1825, which I have reverted in #1845 that led to the tests failing as mentioned above. Sorry

[jamesmkrieger]

Also, please can you check if this fixes the problems with the latest numpy versions too (see #1655)?

The numpy thing with pyparsing is now checked and that indeed fixed it.

[jamesmkrieger]

closed by #1851