I finally checked the OPLS parameters generated by the current version
of moltemplate (and Jason's "oplsaa_moltemplate.py" script):
Conclusion: They are correct.
This is consistent with Arun's tests that he ran in October. (I
cc'd his email to the end of this post)
However, as Connor noticed, the OPLS parameters have changed since
they were first published in 1996. If you want to be precise, you
have to indicate which version of OPLS. Currently the TINKER website
is distributing the the July 2008 version of OPLS, and this is what
moltemplate (and "oplsaa_moltemplate.py") is using.
However Arun's point that improper interactions are not quite
correct is still valid I will post a fix for this in the next few
days.
Thanks for helping fix these issues Connor and Arun!
Andrew
---- details (feel free to skip) ----
During this discussion, we raised two issues with the dihedral
interactions in all-atom OPLS force-field:
1) Conor is right that in many cases, the numbers in the torsion
section of the "oplsaa.prm" file distributed with TINKER often do not
match the parameters from the original 1996 Jorgensen paper on OPLSAA.
This is because, according to comments located in the "oplsaa.prm"
file, these parameters come from the July 2008 version of OPLS (BOSS
version 4.8).
You can download the latest version of these parameters from this web page:
http://zarbi.chem.yale.edu/software.html
http://zarbi.chem.yale.edu/doc/boss49.pdf
2) Sign error?
Fortunately, after looking more carefully, I think there is no sign
error in the dihedral coeffs generated by "oplsaa_moltemplate.py".
(Recall that I thought there might have been a problem with the sign
of the second term in the dihedral interactions in the "oplsaa.lt"
files generated by "oplsaa_moltemplate.py". The V1, V2, V3
coefficients in the oplsaa.prm file located on the TINKER site have
the same sign as the V1, V2, V3 coefficients in the original 1996
Jorgensen paper, and we should interpret them as such. Fortunately,
this is what the "oplsaa_moltemplate.py" script has been doing.)
(I'll stop worrying about the "180" located in the 10th column of the
"torsion" section of "oplsaa.prm". What ever this is, it is not
relevant to us.)
Cheers
Andrew
The dihedral coefficients are correct. I have checked it. The impropers are
the ones that are different in moltemplate. In my opinion you should not
change the position of the central atom.
Thanks again Arun.
I fixed the problem with impropers (it was sort of a pain to
implement). I will finish testing it and post it very soon.