Hi,
Part of the problem is that you are likely dealing with not just two
discrete conformations of the ligand (phenylpropanol by the looks of it) and
protein - there may be 'transition' states occupied below the level of
detection - therefore between the two 'end' states there's probably a
smeared continuum, and the sum of the two discretely modeled occupancies is
not 1 (integral over the smear would be 1 of course but you're not modeling
this). To make things worse, the floppy propanol chain is probably sampling
additional sub-conformations of its own...
Notably, since the ligand is anchored via its phenyl ring, and the
transition between two states does not perturb the apparent occupancy of the
ring atoms (even though in strict sense there are multiple sets of atoms
superposed with each set having unique vibraitonal modes and so forth).
Q1: for the purposes of whatever research you're doing I would argue that
modeling the minor ligand position is acceptable.
Q2: local omit map, kick-omit map, and the density-filling procedure using
partially occupied random atoms followed by refinement (CCP4 post from about
7-8 months ago mentioned that very nice paper which I have since forgotten,
alas) - these methods may give you maps with some more clarity.
Good luck,
Artem
"Nothing is built on stone; all is built on sand, but we must build as if
the sand were stone"
Jorge Luis Borges
-----Original Message-----
From: CCP4 bulletin board [mailto:[log in to unmask]] On Behalf Of Matt
Merski
Sent: Wednesday, September 23, 2009 6:10 PM
To: [log in to unmask]
Subject: [ccp4bb] Modeling Multiple Ligand Conformations
Hi all,
I am solving a series of protein-ligand complex structures in which the
larger ligands typically cause an expansion of the binding site,
changing the receptor into an open conformation, while the smaller
ligands do not change receptor conformation upon binding. In one
structure (1.49 Ang resolution), after several iterations of refinement
with phenix (R-work = 0.1609, R-free = 0.1918), I see both
conformations: all of the closed receptor conformation and the backbone
conformation and 6 out of 7 of the side chains of the open conformation
can be seen in the 2Fo-Fc map (70% occupancy for the closed
conformation, 30% for the open conformation). Modeling in the ligand as
one 100% occupancy pose that fits in the closed receptor conformation
(Fig. 1) gives some negative Fo-Fc density up to sigma = 3.8 and does
not explain the appearance of the open receptor conformation. If I model
the ligand in at 70% occupancy, the negative 2Fo-Fc density is
eliminated and some positive Fo-Fc density appears around the ligand
(Fig.2), suggesting the presence of a second ligand conformation but
there is not enough density to unambiguously place a second ligand
conformation (corresponding to the open receptor conformation). When I
model in a second ligand conformation (Fig. 3), then after refinement
positive Fo-Fc density disappears without any negative Fo-Fc density
appearing but no 2Fo-Fc density appears after refinement to confirm the
correctness of the ligand pose.
Q #1: should I model the ligand in if it eliminates the Fo-Fc positive
density and doesn't cause negative density? Or should I leave it out
despite the indirect evidence from the altered receptor conformation for
this additional pose? Because this is a series of ligands we know that
the larger receptor conformation implies that a ligand is present in a
pose that opens up the binding site but it's difficult to confirm the
second ligand pose.
Q #2: Is there any way to directly test the correctness of the second
pose that is not dependant on the 2Fo-Fc maps? Is there some kind of
statistical test (such as a local R-factor) to show that the pose is not
in direct disagreement with the data?
A pdf of the figures is available at
http://blur.compbio.ucsf.edu/~merski/figs.pdf
Thanks for your help.
Matthew Merski
UCSF Dept. of Pharmaceutical Chemistry
[log in to unmask]
|