Re: [AMBER] Non-integer/zero charges from RESP fitting?

From: David A Case <>
Date: Wed, 13 Jan 2021 08:15:17 -0500

On Tue, Jan 12, 2021, Kenneth Huang wrote:

>The odd thing I've noticed is that while the canonical nucleosides
>themselves add up to zero, my modified NAs are all very slightly off
>(0.000001). My instinct is that it's a rounding error, but when I check the
>output, it says the net charge should be 0,
>> Reading input for molecule 6 weight: 1.000
>> inosine
>> Total charge (ich): 0
>> Number of centers: 30

The "Total charge (ich)" look like an integer, not a real number.

>Which from my limited understanding of it, looks like it should restrict
>the charge to N decimal places (6?) which leaves me puzzled as why it'd be
>unable to zero the charges, instead of leaving them at 0.000001? Or am I
>missing something very obvious?

It's actually quite tricky to take a list of floating point numbers that add
up to zero (to some very small tolerance, like 10^-12 or better), and
convert them into a list that, when printed with an F12.6 format, appear to
add to zero with no roundoff. Ideas are welcome; maybe Dave Cerutti knows
how this problem is handled in mdgx.

If you are making just a limited number of units, the only workaround I know
of is to just hand-edit the output .mol2 or .off files, and force the
charges to add up to exactly zero. Or, just ignore the problem.


AMBER mailing list
Received on Wed Jan 13 2021 - 05:30:03 PST
Custom Search