On 30/08/12 19:47, Seth Harris wrote:
Hi all,
I and others have occasionally been plagued by the "Add Terminal
Residue" function placing the new residue somewhere distant. While
there are various workarounds (reject and try again, do more searches,
real space refine or regularize after it's been placed which kind of
sucks it over to the build point..., workaround that with fix atom and
then clearing the fix atoms...). In my current case of 3-ish Å
resolution and doing a fair bit of building from scratch, residues
I've just built get pulled out of visually obvious density when I do
my old favorite workaround of RSR after the new residue ends up off
screen 20 Å away from the residue I picked to add it to.
I have some suggestions, or else questions if there are settings that
already achieve the below:
1. Would it be possible for "Add Terminal Residue" to only consider
density within a reasonable distance from the selected attachment
point residue?
Not quite what you are asking for, but this is what I use. Perhaps they
should be defaults.
(set-add-terminal-residue-do-post-refine 1)
(set-terminal-residue-do-rigid-body-refine 0)
2. Second area is that at this resolution, or at least with the maps I
have at this resolution, the behavior of real space refinement which
I've grown accustomed to has gone kind of squidgy (hard to describe,
but less snappy and during real space refinement things will drift
away a bit from what visually looks like a good fit. I assume the
lower resolution map has 'softer' edges, perhaps, and have yet to
explore map sharpening, e.g.). I have, however, played with the
refinement weights, but it doesn't much improve the way that side
chain density can stop the main chain from translating into good
position, or sometimes the side chain battles the main chain for its
rightful density. I wondered whether there is or could be a real space
refinement weighting scheme where main chain atoms have a stronger
weight and the further one goes out a side chain the less pull (lower
weight) it would have towards density. I think this might prove quite
useful at the moderate/low resolution range where a fair bit of manual
building transpires. It could prevent long side chains from invading
neighboring strands' main chain density
This was a request from Phil Evans and implemented in 0.7-pre. This
should not happen now - but maybe I've made a mistake.
and be kind of like working with polyala but without having to mutate
away and later put back side chains, and also still having some weight
for the larger side chains that do have good density which are
critical landmarks. It would be neat to be able to tune the fall off
of such a weighting gradient from strong (where side chains are just
about absent in the context of RSR) up to zero fall off, matching
current behavior with all atoms present.
Yes, I have been thinking similar thoughts. Not done anything about
them though :-)
I suppose I could mimic this to see how it works by setting
occupancies to zero (or partial) in all side chain atoms beyond
C-beta, e.g. but I wasn't sure with low occupancy if they still ride
along appropriately with their main chain. Perhaps this parameter
already exists?
It does not.
Paul.
