[AMBER] MBAR with Amber 18

From: Ulf Ryde <ulf.ryde.teokem.lu.se>
Date: Wed, 1 Jul 2020 10:25:00 +0000

Hi!

Are the MBAR energies with Amber 18 really correct?


With Amber 16, the MBAR energy for the clambda value coincided with EPtot.

For example, with clambda=1.00:


MBAR Energy analysis:

Energy at 0.0000 = -10260.7249

...

Energy at 1.0000 = -10304.4021

 ------------------------------------------------------------------------------



| TI region 1



 NSTEP = 500 TIME(PS) = 521.000 TEMP(K) = 306.35 PRESS = 13.8

 Etot = -8817.4742 EKtot = 1486.9279 EPtot = -10304.4021


With Amber 18, this is no longer the case:


MBAR Energy analysis:

Energy at 0.0000 = -9446.2036

...

Energy at 1.0000 = -10361.9441

 ------------------------------------------------------------------------------



| TI region 1



 NSTEP = 500 TIME(PS) = 521.000 TEMP(K) = 299.03 PRESS = -111.7

 Etot = -8831.7112 EKtot = 1451.4229 EPtot = -10283.1341


Moreover, when used with pyMBAR or other software, the results are strange and do not coincide with the TI results (which are instead based on dV/dL from the Amber output file), as they do for well-behaving cases with Amber 16.

In fact, if the MBAR Energy for clambda is subtracted from that of the other lambda values, different snapshots often give exactly the same results for long time intervals, although dV/dL varies.


Best Regards,

Ulf

_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Wed Jul 01 2020 - 03:30:03 PDT
Custom Search