pdb_selaltloc removes residues with (only) alternate locations

[mgiulini]

Describe the bug
pdb_selaltloc removes residues in which all atoms show alternate locations

To Reproduce

grep -rn pdb4xoj.pdb -e "SER A  85"
pdb_selaltloc pdb4xoj.pdb > pdb4xoj-occ.pdb
grep -rn pdb4xoj-occ.pdb -e "SER A  85"

Expected behavior
pdb_selaltloc should not delete the residue but rather pick one of the two available alternate locations

Desktop (please complete the following information):

• OS: Linux
• Python Version 3.9

pdb4xoj.pdb.zip

[joaomcteixeira]

Thanks for this @mgiulini
I will try to address it ASAP.
Cheers,

[mgiulini]

thank you @joaomcteixeira
I just linked the corresponding issue in the artic3d repo

[joaomcteixeira]

Hi, sorry for the delay in addressing this issue. I have been a bit overwhelmed on other fronts. I will address it in the next few weeks. It seems something that needs to be inspected carefully.

[mgiulini]

hi Joao! no worries, I could actually try to give a look into it by myself, as I would like to use pdb_selaltloc in another application, what do you say?

[amjjbonvin]

This is a serious issue that should be solved.

[mgiulini]

I'll look into it!

[joaomcteixeira]

Go for it. Look carefully at the way @JoaoRodrigues and I designed the tests. I suggest you first write the tests for this new case and then try to correct the code for that. Be sure first that the PDB is formatted well. Some issues with PDBs are unsolvable because the PDB itself is wrongly created.
Best,

[mgiulini]

I understood the problem, but I think it would take ages for me to solve it without breaking the code, as the algorithm is very complicated.

the problem occurs when flush_resloc_occ is called

pdb-tools/pdbtools/pdb_selaltloc.py

Lines 302 to 318 in 2a070bb

	def flush_resloc_occ(altloc_lines, **kw):
	"""Flush the captured altloc lines by highest occupancy."""
	# only the selected altloc is yieled
	highest = 0.00
	altloc = ' '
	# detects which altloc identifier has the highest occupancy
	for key, lines2flush in altloc_lines.items():
	# we check only the first line because all atoms in one identifier
	# should have the same occupancy value
	occ = float(lines2flush[0][54:60])
	if occ > highest:
	altloc = key
	highest = occ
	for line2flush in altloc_lines[altloc]:
	yield line2flush[:16] + ' ' + line2flush[17:]

You can clearly see from the code from the code that if the input dictionary altloc_lines has a key with highest possible occupancy (e.g., an atom with no alternate location, occupancy = 1.0), then the function will ignore all the other lines, no matter their occupancy. As an example, residues LYS84 and SER85 of the pdb 4xoj give rise to the following lines

{' ': ['ATOM    536  N   LYS A  84      -4.827  -2.055  19.735  1.00  9.68           N  \n', ... some other coordinates], 
'A': ['ATOM    537  CA ALYS A  84      -4.644  -3.431  19.279  0.70 10.05           C  \n', ... some other coordinates],
'B': ['ATOM    538  CA BLYS A  84      -4.643  -3.451  19.329  0.30 10.14           C  \n', ... some other coordinates]}

according to the function, the first key (' ') is selected and the others are simply ignored.

Possible solutions could be:

• introduce an if clause in this function to always preserve the atoms with 1.0 (key = " "), and check only among the other keys which is the one has the highest occ
• drop the flush_func_multi_residues logic and read the residues sequentially (according to their number)

I tried the second option, but it broke the tests..

[joaomcteixeira]

I would say the first option sounds better. No problem accommodating the if clause.

Also, feel free to adventure yourself rewriting the inner code if you feel so 😈 . As long as the already implemented test cases pass and the API itself is kept. But first, i suggest trying the if-clause to see what happens. Good luck and many thanks

[joaomcteixeira]

Thanks for all this input Marco. Addressing it in steps:

1)

according to the old implementation the residue ILE 25 should be excluded, but in my opinion that's wrong. (in #154 coment)

I revisited the tests, and they are correct to my knowledge on altlocs. ILE 25 is excluded if users don't give any option because it has the lowest occupancy. To select the ILE, the command has a trick; users have to specify the space:

pdb_selaltloc -' ' tests/data/dummy_altloc.pdb

In this way, ILE 25 is selected instead of LEU 25. Btw, occupancy should add up to one, and in the example, dummy numbers are used (0.61 and 0.90).

2)

There are several errors associated with this PDB when using pdb_selaltloc. On the one hand, SER 85 disappears. On the other hand, several atoms from other residues also disappear if we give -A or -B as options.

This PDB has an altloc formatting I have never seen before, that is: Atoms with alternate locations do NOT have the altloc identifier of the default. For example, the default is ' ' (space) for single location atom, while altlocs use directly to A and B and no ' '. I had never seen this before and completely overlooked this possibility when I developed the current version of selaltloc.

My question: Is this a correct altloc formatting? @JoaoRodrigues @amjjbonvin need your help with this.

If yes, I will need to rewrite the script considerably. It's not a problem, but I want to double-check with you before doing it.

Cheers!

[amjjbonvin]

This is a weird one… But if it comes from the PDB then we should ideally handle it.

…

My question: Is this a correct altloc formatting? @JoaoRodrigues <https://github.com/JoaoRodrigues> @amjjbonvin <https://github.com/amjjbonvin> need your help with this.