Hi all,
I am trying to mutate ASP to ASN on a protein using the TI one-step method.
I have a frame from a long simulation and after heating, I am using the
following setup:
##################################################
imin = 0, nstlim = 20000000, irest = 1, ntx = 5, dt = 0.002,
ntt = 3, temp0 = 298.0, gamma_ln = 2.0, ig = -1,
ntc = 2, ntf = 1,
ntb = 2,
cut=12,
ntp = 1, pres0 = 1.0, taup = 2.0,
ioutfm = 1, iwrap = 1,
ntwe = 500, ntwx = 50000, ntpr = 500, ntwr = 500,
icfe = 1, ifsc = 1, clambda = %L%, scalpha = 0.5, scbeta = 12.0,
logdvdl = 0,
ifmbar = 1, mbar_states = %L2%,
mbar_lambda = %L1%,
timask1 = ':34', timask2 = ':130',
scmask1 = ':34', scmask2 = ':130',
/
&ewald
skinnb = 2,
dsum_tol = 1E-06,
/
&wt type = 'END'
/
#########################
windows=$(seq 0.0 0.04 1.0)
#########################
I use "alchemical_analysis" to get free energy, and because the whole
difference that I wanna show is around 1 kcal, I need to have a low
standard deviation. The problem is with increasing the number of lambda to
45 or increasing the simulation time, I never reach to a result with a TI
dhdl graph without a sharp kink, as well as a good overlapping matrix that
is reproducible if I do the simulation one more time, and the difference
between these two simulations with the same setup is bigger than one most
of the time. What can I change? I already increased to the point 0.01
distance between lambda in the sharp point, but it didn't help and there
are also other areas with bad overlapping matrix.
Best
_______________________________________________
AMBER mailing list
AMBER.ambermd.org
http://lists.ambermd.org/mailman/listinfo/amber
Received on Fri Sep 20 2019 - 03:30:02 PDT
