4
$\begingroup$

What are all of the different ways to derive the partial charges, van der Waals interactions, bond lengths, etc. of an amino acid (in other words, all of the parameters that could be used in a molecular mechanics force field or other related method - such as a Poisson-Boltzmann equation solver)? I'm especially interested in the strengths and weaknesses of each method.

Complete fitting protocols would be greatly appreciated, although all I'm really looking for is a few sentences describing each method and its broad strengths and weaknesses.

$\endgroup$
  • 3
    $\begingroup$ Please, don't tell people to list one method per answer. In general, votes are supposed to reflect upon the quality of a user's answer as a whole, rather than serve as a mechanism for evaluating different methods. Also, in general, we try to restrict the number of wikified questions, for reasons that have been discussed on Meta. $\endgroup$ – Geoff Oxberry Mar 20 '12 at 1:08
  • $\begingroup$ @GeoffOxberry This is exactly the kind of question that this community, and possible this community alone, is qualified to answer. However, as can be seen from the people gracious enough to answer already, a complete answer is too much to ask from one person. Additionally, when it comes to fitting methods, users tend to be experts in one or two methods, and then have less knowledge about the rest. It is exactly this expertise that I would like to tap. Is there someplace on meta where I could make this case? $\endgroup$ – tel Mar 24 '12 at 22:48
  • 1
    $\begingroup$ @GeoffOxberry Also, I'm not trying to hold a vote on which method is best, I would simply like to know what methods exist, and more importantly collect some background information on each from people who have actually used them. $\endgroup$ – tel Mar 24 '12 at 22:49
3
$\begingroup$

To ask about all ways is too ask too much, I guess. For the computationally most efficient way to encode the geometric information, see Section 2 of my survey paper Molecular modeling of proteins and mathematical prediction of protein structure, SIAM Rev. 39 (1997), 407-460.

$\endgroup$
3
$\begingroup$

A general answer to the question is just impossible. This answer is limited to modeling the electrostatic term.

We have recently published a detailed comparison of different methods to compute atomic charges, using two sets of 100+ penta-alanine conformers. (link) We tested both the robustness of the charges with respect to conformational changes and the accuracy with which they reproduce the dipole moment (as a simple way of checking the accuracy of the ESP). For the parametrization of a force field, you want a method to assign charges that is successful in both tests.

The Hirshfeld-I method gives the best trade-off between the two tests. However, we also found that for ionic systems (e.g. oxides) Hirshfeld-I overestimates the ESP surrounding isolated molecules, so make sure you double check this for the system you are interested in. (link) If your molecule is rigid, ESP-fitted charges are a better choice because the erratic conformational dependence is no longer an issue.

Anyway, whatever method you use, without including atomic dipoles, one can never parameterize electrostatic interactions up to chemical accuracy, i.e. with errors below 1 kcal mol-1. For rigid molecules, it is possible to make a multipole expansion of each atom-in-molecule density -- with any partitioning scheme such as Hirshfeld-I, Voronio, Mulliken or QTAIM -- to reach any desired level of accuracy for the electrostatic term. (link)

$\endgroup$

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.