Re: [AMBER] deltaG binding obtained from AmberTools 12 and 13 are different

From: Jason Swails <jason.swails.gmail.com>
Date: Tue, 3 Sep 2013 08:05:26 -0400

On Tue, Sep 3, 2013 at 5:04 AM, Xioling Chuang <xioling_c.hotmail.com>wrote:

> Dear Jason,
>
> Sorry for incomplete information. I used MMPBSA.py and I checked the
> energy components. I found that the energy difference obtained from AT12
> and AT13 is vibrational energy (from nmode), not GB energy. The vibrational
> energy were
> ----------------------------------------
> AT12 AT13
> ----------------------------------------
> cpx 540.8 524.6
> receptor 303.3 295.5
> ligand 243.9 247.9
> delta -34.4 -46.9
> ----------------------------------------
>
> The input is
>
> &general
> startframe=1, endframe=1,
> interval=1, keep_files=0,
> verbose=1,
> /
> &gb
> igb=2, saltcon=0.0,
> /
> &nmode
> nmode_igb=1, nmode_istrng=0.0,
> drms=0.1, maxcyc=1000,
> /
>

The drms=0.1 setting is most likely the problem. Normal mode calculations
are highly sensitive to the starting conditions, and must be carried out at
a local minimum to be valid. With such a loose convergence criteria, the
calculation will go faster (since it is easier to converge to the
'requested' limit) but I would not trust the results.

With drms=0.1, my guess is that AmberTools 12 and AmberTools 13 produce
different 'minimized' final structures. This can happen if a some
snapshots took an 'extra' minimization step in AmberTools 13 compared to
AmberTools 12 because the drms happened to be barely satisfied in one
version and barely not satisfied in another (and since you only did one
frame, there's no way for averaging to smooth out a potential irregularity).

My suggestion is to use the default values for drms (0.001) and maxcyc
(10000). If snapshots don't converge, you should increase maxcyc, not
decrease drms. If the normal mode calculations take too long, consider
using either the quasi-harmonic approximation instead, or potentially look
into the truncated normal mode estimation proposed by the Ryde group. [1]

Good luck,
Jason

[1] J Comput Aided Mol Des (2009) 23:6371, DOI 10.1007/s10822-008-9238-z

-- 
Jason M. Swails
BioMaPS,
Rutgers University
Postdoctoral Researcher
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Tue Sep 03 2013 - 05:30:03 PDT
Custom Search