[gmx-users] problems with difference in the COM distances in US sampling

[ABEL Stephane]

Hello 

I am trying to perform some US calculations to obtain the potential of mean force of the insertion of a molecule inside a lipid bilayer. For that I follow the tutorial of Dr Justin Lemkul with some modifications and start the simulations with the following mdp 

pull_group1_name        = MEMB
pull_group2_name        = FEI
pull_ngroups            = 2
pull_ncoords            = 1
pull-nstxout            = 500
pull-nstfout            = 500
pull_print_com1         = yes
pull_print_ref-value    = yes
;pull_print_components   = yes
pull_coord1_type        = umbrella      ; harmonic biasing force
pull_coord1_geometry    = distance      ; simple direction decrease
pull_coord1_groups      = 1 2
pull_coord1_dim         = N N Y
pull_coord1_k             = 1500
;pull_coord1_rate        = -0.01
pull_coord1_start       = yes 

And indeed after 800 ps the molecule move to the bilayer center and  the distance between the COMs of the molecule and the bilayer reported in the pullx file is close 0 nm. 

However when I do the second step of the tutorial with the perl script provide in the  J. lemkul tutorial (perl_distance.pl) to select the US windows, the distance is greater than around 1 for the same config . see below

the pullx = 

800.0000        0.00105624      -4.02734

the COM-distance computed with perl script  after the membrane was centered in the box with -pbc mol is 

800     1.264

Why  this difference, and to choose the good interval for the US sampling

I use the following command trough the perl script 

distance_PMF$ gmx_mpi distance -s ./US_Files_step1/PHE_POPC_Memb_CHARMM_PMF_1500Kj_01rate_US_step1.tpr -f conf800.gro -n System.ndx -oall dist0.xvg -select 'com of group MEMB plus com of group FEI'



Thank you in advance