[AMBER] questions about paramfit

From: ttz <tzhao97.gmail.com>
Date: Thu, 9 Apr 2020 23:54:12 -0500

   Dear Amber experts:

   I am new to parameter fitting area, and have some fundamental questions
   hoping anyone can help with. It may be boring to answer these such
   fundamental questions, so any suggested reference or searching direction
   would be greatly appreciated.

    1. If the optimized bond parameters(for torsion term) are negative, are
       those parameters is reasonable (of chemical meaning )? Dose it means the
       calculation didn’t converged to global minimum and need continuous
       optimization ?

    2. what is the criteria or reference for function value? Like for 27

    3. How many input structure would be good enough for 27 dimensions fitting
       (for torsion term)?

    4. How can I make sure that the optimized parameters is good enough to do
       simulations? Or How to make sure the calculation converged to the global
       minimum? One way I know is to redo the calculation with the same input
       structures and quantum energy data to make sure get the same optimized
       parameters, is that right?

    5. Excepting increase input structures or change the algorithm and other
       function parameters, is there any other suggested way to help the
       calculation converged to global minimum?

    6. For a stiffness ring structure(like conjugated 5 mem ring), can I
       optimize the dihedral within the ring structure? Because when I include
       that dihedral , the fitting cannot converge (calculation up to two
       days). But if I exclude that dihedral, it can easily converge. But the
       function value is too large, Function value with fitted parameters =
       15578.0208, R^2 = 0.9808. so I don’t know the problem is due to the
       bad input data or I just cannot optimize that dihedral

   And attached my fitting input files and output files.

   Thanks in advance!

   Thank you,


   Sent from [1]Mail for Windows 10


   1. https://go.microsoft.com/fwlink/?LinkId=550986

AMBER mailing list

Received on Thu Apr 09 2020 - 22:00:02 PDT
Custom Search