Closed speleo3 closed 4 years ago
ping @josemduarte, is this an issue in biojava as well?
Good catch, thanks @speleo3 ! I could track this back to an issue in BioJava. I will implement a fix that does not add bonds when alt locs are not the default '.' and are different (e.g. if atom 1 is 'A' and atom 2 is 'B').
There's in any case a certain ambiguity in alt loc definition. For instance what should happen when:
@josemduarte Does case 1 exist? Do you have an example?
Case 2 should always link both (., A) and (., B) for what I've seen so far. Do you have an example where that's wrong?
5MOO MET104/N is an example for case 2 where MMTF is missing the bond from ILE103/C to MET104/N-B:
Thanks @speleo3 . I don't have an example of case 1, perhaps it doesn't exist, was simply trying to look at all possible edge cases.
I've submitted a fix for BioJava. For both case 1 and 2 it will always link '.' to A and B.
In any case it will still be some time before the mmtf files are updated.
The issue was fixed in BioJava 5.4.0. The mmtf update pipeline was also updated accordingly. All new mmtf files should have this fix. The old ones will have to be regenerated at some point.
All files have been regenerated now. The issue should be gone from all files now.
The MMTF file for 5MOO has bonds between alternate conformations of residue MET104 (for hydrogens which are only in one conformation). I don't see those bonds in the mmCIF file, so it looks like this is not a "primary-data-issue".