VMD-L Mailing List

From: Jumper, John (John.Jumper_at_DEShawResearch.com)
Date: Thu Feb 26 2009 - 12:21:15 CST

There is a bug in VMD's weighted RMS fits for version 1.8.7a24. The
correlation matrix for the fit is calculated with squared weights while
the
translation vector is calculated with unsquared weights. This results
in an
incorrect fit and a wrong RMSD value.

The bug can be demonstrated by running the following commands on the PDB
file
at the end of this email.

gopython
from atomsel import *
w = [1.0, 1.5, 2.0, 2.5]
s1 = atomsel('index 0 to 3')
s2 = atomsel('index 4 to 7')
s1.rmsd(s2, w)
s1.move(s1.fit(s2, w))
s1.rmsd(s2, w)

The RMSD value before the fit is significantly lower than the value
after the
fit. The following patch to fitrms.c fixes the bug.

--- fitrms.c.bak 2009-02-26 11:40:57.416552000 -0500
+++ fitrms.c 2009-02-26 11:41:25.159180000 -0500
@@ -166,9 +166,9 @@
     double x1 = w * (vv1[0] - t1[0]);
     double y1 = w * (vv1[1] - t1[1]);
     double z1 = w * (vv1[2] - t1[2]);
- double x2 = w * (vv2[0] - t2[0]);
- double y2 = w * (vv2[1] - t2[1]);
- double z2 = w * (vv2[2] - t2[2]);
+ double x2 = (vv2[0] - t2[0]);
+ double y2 = (vv2[1] - t2[1]);
+ double z2 = (vv2[2] - t2[2]);
     aatmp[0] += x2 * x1;
     aatmp[1] += x2 * y1;
     aatmp[2] += x2 * z1;

rmsfit.pdb:

CRYST1 100.000 100.000 100.000 90.00 90.00 90.00 P 1 1
ATOM 1 C A X 1 1.110 -2.448 1.069 1.00 0.00
C
ATOM 2 C A X 1 -6.396 -6.853 10.913 1.00 0.00
C
ATOM 3 C A X 1 9.566 -8.395 7.453 1.00 0.00
C
ATOM 4 C A X 1 10.124 0.427 9.370 1.00 0.00
C
ATOM 5 C A X 1 0.000 0.000 0.000 1.00 0.00
C
ATOM 6 C A X 1 -7.911 -7.880 11.156 1.00 0.00
C
ATOM 7 C A X 1 10.937 -1.364 10.292 1.00 0.00
C
ATOM 8 C A X 1 10.380 -5.561 7.382 1.00 0.00
C
END

Thank you.

John Jumper
D. E. Shaw Research