0000034: Packer won't mutate a post-trans-mod residue

MantisBT - Rosetta
View Issue Details

ID	Project	Category	View Status	Date Submitted	Last Update
0000034	Rosetta	[All Projects] Incorrect Results	public	2012-07-02 15:19	2013-05-13 11:31

Reporter	smlewis
Assigned To	jwillis0720
Priority	normal	Severity	major	Reproducibility	unable to reproduce
Status	assigned	Resolution	open
Platform		OS		OS Version
Product Version
Fixed in Version
Application(s) Affected	all that use resfiles (fixbb promeniently)
Command Line Used	fixbb -resfile resfile_that_mutates -s pdb_with_pre-existing_postransmod.pdb
Developer Options
Fixed in SVN Version

Summary	0000034: Packer won't mutate a post-trans-mod residue
Description	This is ported from an email report from Jordan Willis: Hi, I was wondering if anyone is aware that the resfile command: NUM CHAIN PIKAA <canonical AA> will not replace the AA if the AA you are going to replace, has a post translational modification. I have only discovered this on a PDB that originally contained sulfated tyrosines when I tried to mutate those tyrosines to a canonical AA. Is this a known issue? Or is it an issue? Jordan
Steps To Reproduce
Additional Information	(as yet unconfirmed)
Tags	design, PackerTask, packing, resfile
Relationships
Attached Files

Issue History
Date Modified	Username	Field	Change
2012-07-02 15:19	smlewis	New Issue
2012-07-02 15:19	smlewis	Status	new => assigned
2012-07-02 15:19	smlewis	Assigned To	=> smlewis
2012-07-31 16:07	smlewis	Assigned To	smlewis =>
2012-08-22 16:24	smlewis	Note Added: 0000082
2012-08-22 16:24	smlewis	Assigned To	=> smlewis
2012-08-22 16:24	smlewis	Status	assigned => new
2012-08-29 08:34	smlewis	Assigned To	smlewis =>
2012-09-04 19:32	Labonte	Tag Attached: design
2012-09-04 19:32	Labonte	Tag Attached: PackerTask
2012-09-04 19:32	Labonte	Tag Attached: packing
2012-09-04 19:32	Labonte	Tag Attached: resfile
2013-05-13 11:28	smlewis	Assigned To	=> jwillis0720
2013-05-13 11:28	smlewis	Status	new => assigned
2013-05-13 11:31	smlewis	Note Added: 0000294

Notes

Some notes from the email thread about this:

the problem here is in variant matching.

In this function:
void ResidueLevelTask_::allow_noncanonical_aa(
...
if ( original_residue_type_->variants_match( **aas_iter ) &&

There is a variant matching check. If you start sulfated, ONLY
sulfated variants match and will be allowed. EMPTY does not affect
original_residue_type_ and will have no effect here.

This is perfectly strict variant matching. You require weaker
matching, which ignores mismatches due to *some* variants (sulfated,
etc) and prevents mismatches due to *other* variants (chainbreak,
terminus, etc).

You can try tweaking this check to not care about your subset of
variants: perhaps a flag "allow_design_tofrom_PTMs", and then modify
this check to ignore variant mismatches caused by some curated set of
variants (sulfated, phosphorylated, etc).

You can try removing this check. It will have nasty side effects like
Rosetta will try packing the chainbreak variants or terminus variants
into the middle of the chain. Hopefully the scorefunction will reject
that.

Moving TYR plus S to its own params file as TYS may work. Doug corrects me on the Dunbrack lookup (good). After further
thought, I also realize that Rosetta ignores the PDB standard, in that
it names sulfated tyrosine TYR and adds a sulfate if those atoms are
present, but the PDB uses TYS. So, creating a S-TYR residue type
file, with the name and name3 TYS, would fix your variant matching
problem AND improve the PDB reader to be more standard-compliant.

This may cause the reverse problem, however, where TYR becomes able to
mutate to TYS in all cases. The ResidueLevelTask constructor uses
variant matching just like the troublesome allow_noncanonical_aa
function, and it will let all aa_tyr pass equally (this is how HIS and
HIS_D both work.).

Currently the code distinguishes via "is_protein" and the aa type
being aa_unk - perhaps there is need for a third level of distinction
for chemically modified but not aa_unk residues - they should be
treated with the aa_unk logic.