Unit test for Reduce - Githubissues

russell-taylor commented 3 years ago

There is a test_reduce.py script in the source directory.

Extend or replace it as needed.

Ensure that it works as part of cctbx testing.

russell-taylor commented 3 years ago

This should be modified to test the expected behaviors of Reduce and check against whether the results fall into expected ranges rather than have exact string matches.

russell-taylor commented 3 years ago

The script does not run under Python 3.6. Rebuilding cctbx with Python 2.7 to see if it runs there.

russell-taylor commented 3 years ago

Jane sent some OH tests that we can add to the list.

russell-taylor commented 3 years ago

The script does run under Python 2.7 but it fails for the current master C version for the tests that are run. This may be due to simple reformatting of the USER MOD lines but it is hard to tell from the output.

russell-taylor commented 3 years ago

The 1ssx1H_Ser226-OH_right_140.pdb and new 1ssx1H_Ser226-OH_wrong_17.2.pdb (named 1ssx1H_Ser226-OH_wrong_17.2_Asn220.pdb) files work such that one of them gets flipped and the other does not. There are two problems:

Significantly, the one named "right" gets flipped and the one named "wrong" does not.
Probably insignificantly, the flipped one is 0.03 Angstroms away from the original. This presumably means that we should adjust the difference threshold to be higher.

russell-taylor commented 3 years ago

Looking at why the 1ssx1H_Ser43-OHwrong-60.pdb and right files are not recognized as model files.

Removing the USER MOD lines did not fix this
Removing All of the header lines before the ATOM lines did not fix this
Replacing the header with the one from the Ser226 file that does work did not fix this
Running dos2unix on the original file did not fix this
The broken file appears to have the same columns as the working file

Removing the BARG atoms for the B alternate conformations fixed the problem. These were recorded as interleaved with AARG lines.

russell-taylor commented 3 years ago

The script is able to read 1EX2.pdb, which has A and B alternates in its HETATM records.

russell-taylor commented 3 years ago

The script is able to read the 1ssx.pdb file downloaded directly from the PDB. The problem seems to be the alternate Hydrogen atoms. If I remove them, then there is no problem.

russell-taylor commented 3 years ago

It is just the following line: ATOM 0 H BARG A 141 19,159 27.189 7.221 0.25 3.69 H new

The corresponding A line works and the HA B line following also works:

ATOM      0  H  AARG A 141      18.775  27.301   7.265  0.75  3.69           H   new
ATOM      0  HA AARG A 141      18.805  26.646   4.650  0.75  4.91           H   new

russell-taylor commented 3 years ago

Asked the bulletin board about this to see if it is a bug.

russell-taylor commented 3 years ago

When I remove the B alternate from the model, it flips the Ser43 named "wrong" into the same place as the one named "right".

ReliaSolve / Molprobity2

Unit test for Reduce #90