From: Matthew Ralph Adendorff (mraden_at_mit.edu)
Date: Sun May 10 2015 - 08:44:08 CDT
I am processing a 64-window FEP calculation, 50,000 steps equilibration (100 ps) and 450,000 steps FEP (0.9 ns) data collection (total steps 500,000). My forward and backward runs used a lambda spacing of +/- 0.015625 to generate the 64 windows. Energy values are written every 10 steps (frequent I know, though I'm still in the testing phase).
When I process the alchemical output files in ParseFEP, the program detects only 63 windows and then uses 50,793 effective samples rather than the 50,000 I would have expected. The additional 793 samples are most likely due to the 64th window's 50,000 samples being split across 63 windows. Plotting the free energy results shows that the two runs are offset from each other, with both runs terminating one step too early, i.e. the forward doesn't reach 1 and the backward doesn't reach 0.
When I extract the comment lines from the alchemical output files, the correct values of lambda and lambda2 are being set to facilitate reaching 0 and 1. Within ParseFEP I have used a 6th order G-C and I am using the BAR-estimator.
Any advice would be greatly appreciated.
Laboratory for Computational Biology and Biophysics
Department of Biological Engineering
Massachusetts Institute of Technology
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2015 - 23:21:52 CST