Page 1 of 1
surface tension factor
Posted: Mon May 02, 2011 2:43 pm
by dark_angel83
Dear openmm developers,
I'm using OpenMM with the implicit solvent.
I compared the AMBER implementation with OpenMM, I guess I found a discrepancy: in Amber the nonpolar term is calculated multipling the SAS with a surface tension factor:
Enp = surften*SA. The default value for surften is 0.005 kcal/mol/A2 (page 58).
In OpenMM I don't see this kind of multiplication in OpenMM. Is It a bug or has been managed in some way?
Francesco
RE: surface tension factor
Posted: Mon May 02, 2011 3:22 pm
by friedrim
Hi Francesco,
The code for the nonpolar term used for implicit solvent calculations is carried out for the Reference platform in CpuImplicitSolvent::computeAceNonPolarForce() in the file platforms/reference/src/gbsa/CpuImplicitSolvent.cpp and for the Cuda platform in kReduceObcGbsaBornForces_kernel in the file platforms/cuda/src/kernels/kForces.cu
The calculation of the nopolar term is based on the ACE approximation from the paper:
M. Schaefer, C. Bartels and M. Karplus, "Solution Conformations
and Thermodynamics of Structured Peptides: Molecular Dynamics
Simulation with an Implicit Solvation Model", J. Mol. Biol.,
284, 835-848 (1998) (ACE Method)
The original equation includes the factor (atomicRadii[atomI]/bornRadii[atomI]) to the first power,
whereas in OpenMM the ratio is raised to the sixth power: (atomicRadii[atomI]/bornRadii[atomI])**6
This modification was made in Tinker by Jay Ponder who observed it gave better correlations w/
observed values. He did not think it was important enough to write up, so there is
no paper to cite. The OpenMM GBSA/OBC algorithm is based on the Tinker code.