search

TinkerTools / tinker

Tinker: Software Tools for Molecular Design
https://dasher.wustl.edu/tinker/
Other
129 stars 61 forks source link

MM3PRO Hessian #74

Closed NF-AMU closed 3 years ago

NF-AMU commented 3 years ago

Hi, While using the enkephalin example which is based on the MM3pro ff, I've experienced that the numerical and analytical Hessian elements can differ much. For instance: Comparison of Analytical and Numerical Hessian Elements : 1st Atom 2nd Atom Analytical Numerical Difference ... 16 (Y) 16 (Y) 5.8317 -14.3572 20.1889 16 (Y) 16 (Z) 17.6193 21.9278 -4.3085 16 (Y) 23 (X) 51.6569 46.2261 5.4308 16 (Y) 23 (Y) 45.3146 57.9164 -12.6018 16 (Y) 23 (Z) -0.5540 -3.0916 2.5375 ... in a calculation with van der Waals interactions only.

jayponder commented 3 years ago

You are correct. The Tinker code does not have the exact analytical Hessian for the vdw term in the MM3PRO force field, as MM3PRO uses an odd formulation including a hydrogen bond term for which we do not have the analytical Hessian. The H-bond modification makes the correct code for the analytical Hessian really nasty. The "analytical" result printed by the TESTHESS program uses the standard MM3 vdw Hessian, without the hydrogen bonding modification. The numerical Hessian should be the "correct" Hessian value for MM3PRO.

I'm not sure we will fix this anytime soon, and I'm going to go ahead and close the issue. What I might do is print a warning for the MM3PRO case that the "analytical" Hessian is only approximate. The MM3PRO force field is essentially obsolete anyway, and is really only included in Tinker for historical reasons. If you want to model proteins you should switch to one the standard "biomolecular" force fields such as CHARMM, Amber or OPLS. Or you could use our polarizable AMOEBA force field if you want a more advanced, but slower, model.