Improper chain mapping for protein-RNA complexes causing index error

[zrqiao]

Thanks for creating this great tool!

For some protein/nucleic-acid complex targets (example) we observed IndexError: tuple index out of range similar to what reported in #48 .

Tracking the issue down, for these cases it was caused by oversensitive sequence alignment between the protein and the RNA chain, leading DockQ's chain mapping module to treat two distinct chains as the same. Printing out the alignment result from https://github.com/bjornwallner/DockQ/blob/master/src/DockQ/DockQ.py#L718:

{'seqA': '----------------------------------------------------A-----------A---------------------TC---------CA----------------------------AA-------------------------------------------------------------', 'matches': '                                                    |           |                     ||         ||                            ||                                                             ', 'seqB': 'RHSMDPATFTFNFNNEPWVRGRHETYLCYEVERMHNDTWVKLAQRRGFLANQAKHKHGFLEGRHAELCFLDVIPFWKLDLDQDYRVTCFTSWSPCFSCAQEMAKFISKNKHVSLCIKTARIYDDKGRAAEGLRTLAEAGAKISIMTYSEFKHCWDTFVDHQGAPFQPWDGLDEHSQDLSGRLRAILQNQE'}

A more robust algorithm with gap penalty can be used to alleviate these cases. Is there a plan to improve the chain mapping algorithm? Happy to make a PR to help.

[clami66]

The best way to fix this, I think, is to only align molecules of the same type (i.e. do not align proteins to NAs). Unfortunately I've been quite busy to do it.

But if you have time to fix it yourself (even by modifying the alignment function) then please go ahead and do a PR