Advanced

Comparison of the Efficiency of the LIE and MM/GBSA Methods to Calculate Ligand-Binding Energies

Genheden, Samuel LU and Ryde, Ulf LU (2011) In Journal of Chemical Theory and Computation 7(11). p.3768-3778
Abstract
We have evaluated the efficiency of two popular end-point methods to calculate ligand-binding free energies, LIE (linear interaction energy) and MM/GBSA (molecular mechanics with generalized Born surface-area solvation), i.e. the computational effort needed to obtain estimates of a similar precision. As a test case, we use the binding of seven biotin analogues to avidin. The energy terms used by MM/GBSA and LIE exhibit a similar correlation time (similar to 5 ps), and the equilibration time seems also to be similar, although it varies much between the various ligands. The results show that the LIE method is more effective than MM/GBSA, by a factor of 2-7 for a truncated spherical system with a radius of 26 angstrom and by a factor of... (More)
We have evaluated the efficiency of two popular end-point methods to calculate ligand-binding free energies, LIE (linear interaction energy) and MM/GBSA (molecular mechanics with generalized Born surface-area solvation), i.e. the computational effort needed to obtain estimates of a similar precision. As a test case, we use the binding of seven biotin analogues to avidin. The energy terms used by MM/GBSA and LIE exhibit a similar correlation time (similar to 5 ps), and the equilibration time seems also to be similar, although it varies much between the various ligands. The results show that the LIE method is more effective than MM/GBSA, by a factor of 2-7 for a truncated spherical system with a radius of 26 angstrom and by a factor of 1.0-2.4 for the full avidin tetramer (radius 47 angstrom). The reason for this is the cost for the MM/GBSA entropy calculations, which more than compensates for the extra simulation of the free ligand in LIE. On the other hand, LIE requires that the protein is neutralized, whereas MM/GBSA has no such requirements. Our results indicate that both the truncation and neutralization of the proteins may slow the convergence and emphasize small differences in the calculations, e.g., differences between the four subunits in avidin. Moreover, LIE cannot take advantage of the fact that avidin is a tetramer. For this test case, LIE gives poor results with the standard parametrization, but after optimizing the scaling factor of the van der Waals terms, reasonable binding affinities can be obtained, although MM/GBSA still gives a significantly better predictive index and correlation to the experimental affinities. (Less)
Please use this url to cite or link to this publication:
author
organization
publishing date
type
Contribution to journal
publication status
published
subject
in
Journal of Chemical Theory and Computation
volume
7
issue
11
pages
3768 - 3778
publisher
The American Chemical Society
external identifiers
  • wos:000296597300032
  • scopus:80755155952
ISSN
1549-9618
DOI
10.1021/ct200163c
language
English
LU publication?
yes
id
75d4a8ea-ee64-4118-b030-6f4916ce29ce (old id 2253249)
date added to LUP
2011-12-21 14:59:23
date last changed
2017-07-09 03:11:39
@article{75d4a8ea-ee64-4118-b030-6f4916ce29ce,
  abstract     = {We have evaluated the efficiency of two popular end-point methods to calculate ligand-binding free energies, LIE (linear interaction energy) and MM/GBSA (molecular mechanics with generalized Born surface-area solvation), i.e. the computational effort needed to obtain estimates of a similar precision. As a test case, we use the binding of seven biotin analogues to avidin. The energy terms used by MM/GBSA and LIE exhibit a similar correlation time (similar to 5 ps), and the equilibration time seems also to be similar, although it varies much between the various ligands. The results show that the LIE method is more effective than MM/GBSA, by a factor of 2-7 for a truncated spherical system with a radius of 26 angstrom and by a factor of 1.0-2.4 for the full avidin tetramer (radius 47 angstrom). The reason for this is the cost for the MM/GBSA entropy calculations, which more than compensates for the extra simulation of the free ligand in LIE. On the other hand, LIE requires that the protein is neutralized, whereas MM/GBSA has no such requirements. Our results indicate that both the truncation and neutralization of the proteins may slow the convergence and emphasize small differences in the calculations, e.g., differences between the four subunits in avidin. Moreover, LIE cannot take advantage of the fact that avidin is a tetramer. For this test case, LIE gives poor results with the standard parametrization, but after optimizing the scaling factor of the van der Waals terms, reasonable binding affinities can be obtained, although MM/GBSA still gives a significantly better predictive index and correlation to the experimental affinities.},
  author       = {Genheden, Samuel and Ryde, Ulf},
  issn         = {1549-9618},
  language     = {eng},
  number       = {11},
  pages        = {3768--3778},
  publisher    = {The American Chemical Society},
  series       = {Journal of Chemical Theory and Computation},
  title        = {Comparison of the Efficiency of the LIE and MM/GBSA Methods to Calculate Ligand-Binding Energies},
  url          = {http://dx.doi.org/10.1021/ct200163c},
  volume       = {7},
  year         = {2011},
}