Logs:parameterization

From biowiki
Revision as of 23:11, 21 April 2025 by Eew947 (talk | contribs) (Copied, haven't checked)
(diff) ← Older revision | Latest revision (diff) | Newer revision → (diff)
Jump to navigation Jump to search

The external webpage that describes the usage of poltype.py is on water.

Issues/Feature Requests

OP Bend

  • Differences between the opbend terms from poltype and amides from amoeba '09 parameters
  • Main difference is for CO - N
  • First number is from poltype. Numbers in parantheses are from AMOEBA '09 small molecules and AMOEBA_BIO '09)
# CH3 CO
opbend    401   403     0     0    14.402 (same, N/A)
# CO N
opbend    403   404     0     0    71.940 (12.9, 70.5)
# N CO
opbend    404   403     0     0   107.910 (same, same)
# O CO
opbend    407   403     0     0   122.298 (46.8, 54.0)
# H N
opbend    408   404     0     0     5.755 ( 5.8, 12.9)

Updated polarizabilities

The third column is polarizablity; and the fourth one is vdW, if different with current amoeba.


 O- (in COO-)
 1.600
 3.700/0.1290

 S- (in R-S-)
 4.000
 4.200/0.3550

 N(aromatic HIS)
 1.500

C(aromatic HIS & COO-)
 1.750

H(aromatic HIS)
 0.696


Parameterization

  • Bisector: set XZ component of quadrupole to 0
  • Use parameters from /home/pren/AMOEBA
    • amoeba_v2_new.prm
    • amoeba_2009_ver4.prm.isoprop
  • If all opbend angles for atom X of minimized structure > 4 degrees, remove opbend parameters and set equilibrium angles around atom to 110 degrees
    • Confirm with Jenny if "4 degrees" and "110 degrees" are correct.
  • Set opbend force constant according to amoeba.prm
    • 1.7 for carbonyl C=O, C-C(O), and -N-C=O (amide)
    • ~0.6-0.7 for alkane C?
    • Add opbend to nitromethane-like groups
  • Remove pi-torsion param for C-C=C-C and increase torsion to 5.5 from 4.5
  • quadrupole scaling on OH in alcohol (-Csp3-OH)
    • Use GDMA Input
    • Scale by 70% if gdma v2.2 procedure is used
    • Scale by 60% if gdma v1.2 or 1.3 is used (switch=0 and H radius 0.65)
  • Classify terminal atoms that are the same element together
    • Check Gaussian MBO (bond order to get accurate valence info)
  • Torsion
    • Minimize with HF/6-31G8 or M06L
    • SP using DFT with ML06 functional (in G09)
    • Methyl group torsion 0.299
    • Only single bonds with heavy atoms
  • Plot phase angle without torsion parameters
  • DFT vs HF optimization? Which is better?
  • Modify polarizability for imidazole N and C (based on Jenny's findings)
  • New functional groups
    • fit_struct (run with MP2/6-31++G(2d,2p)
      • bond, angle, opbend
    • polarizability (pol-eps = 0.000001)
    • vdw (compare distance, dimer/interaction energy)
      • To calculate dimer energy from Gaussian output:
        • BSSE Corrected Dimer energy = (1 - 2 - 3) * 621.5095 # Constant to convert HF to kcal/mol
        • Non-BSSE Corrected Dimer energy = (1 - 4 - 5) * 621.5095
        • Values 1, 2, 3, 4, 5 are taken by grepping EUMP2 from Gaussian logfile
        • Take the EUMP2 values from the 5 lines
        • Compare BSSE corrected and non-corrected
        • Try to get somewhere in between
  • Add option for Freq for DMA QM
  • Fix QM torsion for all heavy atoms
  • Generate torsion conformations with AMOEBA rather than QM (add option)

Status

  • What value to use for alkane? 0.6 or 0.7?
  • Question: Amide opbend values don't match.
    • /home/pren/AMOEBA/amoeba_v2_new.prm
    • Values we discussed 1.7

Validation

  • Calculate molecular polarizability: B3LYP/6-311++G(2d,2p)
  • Water-compound Dimer Energy
    • Opt: MP2/6-31++G(2d,2p)
    • SP: MP2/aug-cc-pVTZ & Counterpoise 1,1,1,1,1,1 (using Q-chem/RI TRIM)
    • Use to modify some vdW parameters
    • nitro - water: find O -- H distance, and angle
  • Run QM of nitro as N-O bond distance changes
    • Explore changes in multipoles
  • Run QM with Q-Chem (rimp2)

Bugs/Feature request

  • Convert Tinker 5 to Tinker 4

Paper #1

Method

  • Parametrization: Run HF/6-31G* opt, single point MP2/6-311G** -> GDMA (polgrp, average/symm) -> ESP of  MP2/6-311++G(2d2p)
  • Assign all other valence, vdw, polarizability, ...
  • High-level QM data for dimers/complexes (RI-MP2/aug-cc-pvtz -pvdz)

Results

DFT vs. MP2

  • Molecule/water dimer energy
  • Torsion energy
  • Geometry?
  • (w/ ~10-20 molecules)

AMOEBA vs. DFT

  • Based on DFT vs. MP2 findings

Monomer

  • Vibrational frequency
  • Dipole moment (using different conformations)
  • ESP
  • Geometry (RMSD)

Hetero Dimer w/ water

  • Energy
  • Structure
  • QM:MP/6-31+G* opt and MP2/aug-cc-pvtz for dimerization energy all the *other (validation): DFT/B3LYP 6-31G* opt and dimer energy
  • Correlation w/ solvation energy

Solvation Free Energy

  • Choose 5/20 molecules which should agree with fixed-charge FF and some that should showcase strenght of AMOEBA FF.

Done

  • Implement symmetry class assignment in Python
  • Implement atom type assignment based on SMARTS string
  • Implemented opbend framework (Jwu 05:05, 29 June 2010 (UTC))
    • Added opbend values from amoeba_v2_new.prm (Jwu 22:09, 9 July 2010 (UTC))
    • Re-implemented opbend to use values (Jwu 22:10, 6 July 2010 (UTC))
  • Match polarization group class ID with multipole class IDs
  • print/check bond, angle and torsion deviations (Jwu 03:58, 12 March 2010 (UTC))
  • select ~ 200 molecules
  • From Rizzo Supplementary Info:
    1. 1040
    2. 1014
    3. 1018
    4. 1019
  • Describe Babel instructions (Jwu 22:31, 9 March 2010 (UTC))
  • Remove duplicate multipoles (Jwu 22:33, 10 March 2010 (UTC))
  • Geometry: RMSD AMOEBA vs. QM (Jwu 03:57, 12 March 2010 (UTC))
  • Check polarization groups (Jwu 07:40, 12 March 2010 (UTC))
    • Alcohol (-C-OH)
    • (-C-NH2)
    • Benzene ring w/ CH2 (-CH2-c1ccccc1)
    • Phenol (-c1ccccc1(-OH)
    • (-CH2-C(=O)-NH2)
  • Check electrostatics (Jwu 20:25, 12 March 2010 (UTC))
    • net charges (0 or -1, +1...)
    • dipole and quadrupole
    • ESP
  • minimize your molecule and compare to the Qm structure. RMSD should be small (Jwu 20:25, 12 March 2010 (UTC))
  • Run QM for set of molecules (Jwu 16:41, 17 March 2010 (UTC))
    • Copy correct SDF files and generate load-balanced scripts to run QM
    • Modify poltype to run QM only (Jwu 20:20, 15 March 2010 (UTC))
    • Install babel 2.2.3 and python bindings on TI3d cluster (ddfe) (Jwu 20:20, 15 March 2010 (UTC))
  • QM method (Jwu 21:01, 21 June 2010 (UTC))
    • Perform GDMA on MP2/6-311G** to get charge
    • minimize ESP RMSD to MP2/6-311++G(2d,2p) (or Aug-CC-pVTZ) to get dipole, quadropole
  • Check for amide (OCN), acetic (OCC) to recognize polarization groups
  • Scale amine quadrupoles (Jwu 03:51, 1 December 2011 (UTC))
  • Use ESP convergence (0.5) (Jwu 03:51, 1 December 2011 (UTC))
  • Polarization group of functional group should include adjacent SP3 carbon (Jwu 03:51, 1 December 2011 (UTC))
  • Explain how aromatic carbons are found in Babel on water site (Jwu 04:51, 1 December 2011 (UTC))
  • Explain how charge multiplicity can be changed in SDF file on water site (Jwu 04:51, 1 December 2011 (UTC))
  • Explain in docs, what are from database (and how it can be expanded), and what will be computed. (Jwu 05:07, 1 December 2011 (UTC))
    • How are bond orders determined
      • single vs double bonds
      • aromatic C
      • What are aromatic C information for?

Notes/Misc.

  • In amoebapro4.prm, header “opbendunit 0.02191418”, so actual force constant is “smaller” than amobeabiov4.prm
  • amobeabiov4.prm (doesn’t have the opbendunit in the header)
    • Uses the default: 1/radian^2, where radian =57.29
    • opbend_force_in_prm= actual_opbend_force_used* (radian**2/opbunit)
    • opbendunit is 0.02191418 instead of 1/57.3^2 (when not defined the opbendunit in the header)
    • Force constant in prm should be smaller by 71.9
    • Can’t mix parameters from the amoebapro4.prm and amoebbiov4.prm right now
    • amber handles amoebapro4.prm correctly, but need to check the amobeabiov4.prm

Paper #2

Solvation

  • HFE energy >50 molecules?
  • Solvation energy too negative
    • 23 dimethylpyridine
    • NN dimethylformamide
    • diisopropylamine
    • 4 methoxyacetophenone
    • 2 ethoxyethanol
    • propanone
      • Test with ethanone?
    • Notes in Dropbox:small_mol_johnny\solvation\solv-free-energy.xls

Paper #3, etc.

Solvation with octane (LogP)

Links

Retrieved from "https://biomolmd.org/mw/index.php?title=Logs:parameterization&oldid=2096"