a few öre's worth from sweden ...
there is a surprising number of deposited ligand structures where
re-refinement shows that the ligand is built into noise.
nope, not surprised at all:
- http://xray.bmc.uu.se/cgi-bin/gerard/reprint_mailer.pl?pref=89
- http://xray.bmc.uu.se/cgi-bin/gerard/reprint_mailer.pl?pref=87
- http://xray.bmc.uu.se/cgi-bin/gerard/reprint_mailer.pl?pref=82
- http://xray.bmc.uu.se/cgi-bin/gerard/reprint_mailer.pl?pref=70
- http://xray.bmc.uu.se/cgi-bin/gerard/reprint_mailer.pl?pref=69
i could go on
Would it be fair to say that in the case of a ligand structure, isomorphous
with the native structure but where different R-free sets have been used, we
cannot use the R-free to validate refinement of the ligand? My
how would you use Rfree to "validate refinement of the ligand"? unless it is
megarich in electrons it is not going to make much of a dent in Rfree. use
other criteria to validate your ligand such as (in no particular order):
- convincing fit to the density? (without needing to "explain away" things)
- sensible interactions? (H-bonds, salt links, hydrophobic, no bad contacts)
- good geometry and stereo-chemistry? (bonds, angles, planarity, chirality)
- have we seen this conformation before? (if it occurs in the PDB already; use
valligurl [http://xray.bmc.uu.se/cgi-bin/gerard/reprint_mailer.pl?pref=85] or
MSD/PDBe tools)
- does it explain the biology/binding data/...?
understanding is that it would be contaminated by the phases brought over
with the model from the native data set, and is no longer independent, so
any ligand density could just arise from over fitting.
sorry, but that is obviously nonsense. your unbound protein did not contain
your ligand (the word "unbound" gives it away, really) so its model cannot
possibly contain information that would somehow induce lovely ligand density.
if anything, it should introduce model bias towards the protein because that's
what you use to calculate phases. (by the by, "over-fitting" [refining more
parameters than the information content of the data warrants] doesn't have
much to do with this discussion, so i assumed you meant "model bias")
Does the community still think it is valid to use a slow cool to "reset" the
free R?
far be it from me to presume to represent "the community", of course, but
"yes". i suspect that some of the "holier-than-me" (sic! :-) test-set
worrywarts (use refinement programs that) don't (like to) do SA
--dvd
disclaimer: it's late.
******************************************************************
Gerard J. Kleywegt
[Research Fellow of the Royal Swedish Academy of Sciences]
Dept. of Cell & Molecular Biology University of Uppsala
Biomedical Centre Box 596
SE-751 24 Uppsala SWEDEN
http://xray.bmc.uu.se/gerard/ mailto:[email protected]
******************************************************************
The opinions in this message are fictional. Any similarity
to actual opinions, living or dead, is purely coincidental.
******************************************************************
