MantisBT - Rosetta
View Issue Details
0000013Rosetta[All Projects] Incorrect Resultspublic2012-06-20 16:572012-07-02 14:12
dimaio 
AndrewLeaverFay 
lowminoralways
assignedopen 
All platformsAnyAny
 
 
potentially anything that scores centroid structures
score_jd2.linuxgccrelease -score:weights my.wts -s anything.pdb -in:file:centroid_inputs
0000013: env and interaction graph radii
When centroid structures are scored with env energy alone, the value they return is very different than the env value when scored with both env and, say, vdw. The reason seems to be that -- when scoring with env alone -- the neighbor counts are much lower than they should be.

It seems the problem might be core/scoring/EnvPairPotential.cc:compute_centroid_environment() ... this function is looping over the energy_graph and it looks like the interaction radius is set based on what the SR two-body terms need (via atomic_interaction_cutoff).
create a weights file 'my.wts' that contains:
env 1.0

Then score some pose (using -in:file:centroid_input).

Now edit my.wts:
env 1.0
vdw 1.0

Rescore the same pose. The env value returned in the .sc file is different:
SCORE: score env linear_chainbreak overlap_chainbreak time description
SCORE: 33.471 33.471 0.000 0.000 0.000 3etu_0001
---
SCORE: score vdw env linear_chainbreak overlap_chainbreak time description
SCORE: -75.600 0.234 -75.835 0.000 0.000 0.000 3etu_0001

No tags attached.
Issue History
2012-06-20 16:57dimaioNew Issue
2012-07-02 14:12AndrewLeaverFayAssigned To => AndrewLeaverFay
2012-07-02 14:12AndrewLeaverFayStatusnew => assigned

There are no notes attached to this issue.