[MMTK] getting hessian matrix (and equation) and force vector of a protein conformaion

Konrad Hinsen research at khinsen.fastmail.net
Thu Jan 12 17:17:06 UTC 2012


Hyuntae Na writes:

 > The above equations are the effect between only atom-i and atom-j (relative
 > effect of atom-i by atom-j), so it is only about the off-diagonal. I agree
 > that, for the diagonal of the hessian matrix, the summation of the 2nd
 > derivatives of potential energies should be used.
 > 
 > Does my above approach correct?

Hard to say without verifying all the calculations. It looks OK in
principle, but I didn't check the details.

 > using Conjugate-Gradient and Steepest-Descent. Can I add my own
 > force-field with giving few constraints to several few atoms
 > without touching c source (but within only python part)? Can you
 > give me a guide to do it?

You can easily add force field terms. Under Examples/Forcefield there
are two simple examples. Each of them has an implementation in C, one
in pure Python, and one in Cython (which is a hybrid language between
C and Python). This shows that you can avoid touching any C source,
and even writing your own. However, the pure Python versions are quite
a bit slower, which may or may not be a problem depending on what your
application is.

If you want to write a complete force field similar to Amber, expect
to spend much more time because protein force fields are rather
complicated. But you can definitely do it without touching any of
MMTK's source code, and if you want in pure Python.

If you want to implement a force field that is similar to Amber, in
the sense tht the functional form of the basic terms are the same, you
can also base your force field on the generic MMForceField, just like
MMTK's implementation of Amber does. In that case, all you have to
write yourself is the assignment of parameters to all atoms. This can
still be quite complicated, but then again, that's because those force
fields are complicated.

Konrad.
-- 
---------------------------------------------------------------------
Konrad Hinsen
Centre de Biophysique Moléculaire, CNRS Orléans
Synchrotron Soleil - Division Expériences
Saint Aubin - BP 48
91192 Gif sur Yvette Cedex, France
Tel. +33-1 69 35 97 15
E-Mail: research AT khinsen DOT fastmail DOT net
http://dirac.cnrs-orleans.fr/~hinsen/
---------------------------------------------------------------------



More information about the mmtk mailing list