[gmx-users] Re: How to calculate hydration of lipid headgroups by minimum distance means

Ioannis Beis ibeis at mail.student.oulu.fi
Mon Jan 30 17:41:29 CET 2012

> Message: 6
> Date: Mon, 30 Jan 2012 23:12:46 +1100
> From: Mark Abraham <Mark.Abraham at anu.edu.au>
> Subject: Re: [gmx-users] How to calculate hydration of lipid
> 	headgroups by	minimum distance means
> To: Discussion list for GROMACS users <gmx-users at gromacs.org>
> Message-ID: <4F26893E.2070404 at anu.edu.au>
> Content-Type: text/plain; charset=ISO-8859-7; format=flowed

Hi again,

> On 30/01/2012 10:44 PM, Ioannis Beis wrote:
>> Dear Gromacs users,
>> I am trying to calculate the hydration of lipid headgroups in a lipid
>> bilayer system. More specifically, I would like to find the
>> time-averaged distribution of water molecules around lipid headgroups
>> by means of the closest distance of the water oxygen from any atom of
>> any lipid headgroup and plot this distribution as a function of
>> distance. I would subsequently like to integrate the number of
>> molecules until a distance that I will choose based on the results of
>> the first calculation.
> You can have either the distribution over time of the minimum distance
> between any water oxygen and any lipid headgroup atom from g_mindist, or
> the distribution over space of water oxygen atoms around lipid headgroup
> atoms (a.k.a. radial distribution function) from g_rdf. The above does
> not make clear to me that you are clear about which one you want.
>> g_mindist seems to provide the possibility of doing the latter with -d
>> and also allows counting each molecule in the integration only once
>> with -group. However, what it can do in the first place is essentially
>> different. It calculates the minimum distance among any pair of groups
>> of atoms for all times. With -or it calculates the closest distance
>> among all frames for every residue. So for my system those would give
>> the smallest distance of any water to any headgroup vs time and the
>> closest distance of every water with any headgroup among all frames
>> respectively. This in practice produces something more than twice the
>> van der Waals radius of an atom with small fluctuations.
> Yep. That's what g_mindist does. I think you want g_rdf.

I experimentised with g_rdf. I provide an index with first group all
the headgroup atoms and second group all the water oxygen atoms and
issue the command with either -surf mol or -surf res. The two
hydration peaks appear tiny and the function grows all the way till
around 3 nm (which depends on my box size). This is almost the same as
when I use all the atoms of a single headgroup as a first group. Also,
the manual mentions that -surf doesn't give normalized results, but I
guess this means that it does not divide the time-averaged number of
waters at each slice with the number of waters before each of those
slices and multiply with the volume ratio of each slice over its inner
sphere. If waters in this process would be counted only once each,
this does not justify an average of about 1000 waters per bin in more
than 1500 bins (I just used the default bins) in a system that
contains 6070 water molecules.

So how is the minimum distance between two groups interpreted by the
program with -surf? I would like for each water oxygen the distance
between itself and the closest atom of any headgroup and each water
counted only once. Then I would like to have the time-averaged
distribution defined like this, which should give rise to the
hydration shells, followed by a plateau with lower values than the
peaks and then drop to zero. Could you explain me what g_rdf
calculated in my case and how I could obtain what I am looking for?

Thank you very much in advance!


> Mark
>> g_dist doesn't even calculate minimum distances and g_bond isn't
>> relevant.
>> Is there a way to produce the data I need to use as input for
>> g_analyze so that I can eventually get the distribution with a
>> combination of conventional Gromacs tools? I would appreciate any
>> instructive comment related to my problem.
>> Thank you very much in advance!
>> Ioannis

More information about the gromacs.org_gmx-users mailing list