From: Jerome Henin (jhenin_at_cmm.chem.upenn.edu)
Date: Tue Nov 14 2006 - 08:21:38 CST
Dear Grzegorz,
On Tuesday 14 November 2006 07:40, Grzegorz Jezierski wrote:
> Dear users,
>
> I plan to perform some free energy calculations on systems involving a
> short DNA fragment complexed/bound to a chromophore molecule. I would
> like to use ABF method (in NAMD), as an alternative to umbrella
> sampling. Before, I tried to reproduce the results on
> acetate-guanidinium association, published in J Chem Phys. (2004),
> 121(7); p.2904.
>
> 1.
> A 2-ns ABF simulation (Xi= 4.0 --> 12.0 A) yielded strikingly different
> free energy values than those in the cited paper. I tried to keep as close
> as possible to the original simulation parameters.
A part of the problem is that although the JCP 2004 paper states that the
acetate-guanidinium system is restrained to the C2v geometry, the exact
detail of the restraints is not provided.
They were:
{
angleA1 { angle {A 1 CA} {G 1 N3} {G 1 N1} 40.0 30.0 }
angleA2 { angle {A 1 CA} {G 1 N3} {G 1 N2} 40.0 30.0 }
angleB1 { angle {A 1 O1} {A 1 CA} {G 1 N3} 40.0 27.8 }
angleB2 { angle {A 1 O2} {A 1 CA} {G 1 N3} 40.0 27.8 }
diheA1 { dihe { A 1 O1 } { A 1 CA } { A 1 O2 } { G 1 N3 } 40.0 0.0 }
diheA2 { dihe { A 1 O2 } { A 1 CA } { A 1 O1 } { G 1 N3 } 40.0 0.0 }
diheA3 { dihe { A 1 CA } { G 1 N1 } { G 1 N3 } { G 1 N2 } 40.0 0.0 }
diheA4 { dihe { A 1 CA } { G 1 N2 } { G 1 N3 } { G 1 N1 } 40.0 0.0 }
diheB1 { dihe { A 1 CA } { A 1 O1 } { G 1 N1 } { G 1 N3 } 40.0 0.0 }
diheB2 { dihe { A 1 CA } { A 1 O2 } { G 1 N2 } { G 1 N3 } 40.0 0.0 }
}
As you can see, this set of restraints is more complete and tighter that what
you used. This most probably accounts for the difference in the resulting
PMFs.
> Apart from Langevin
> dynamics (LD), a 2-ns 'normal' (i.e. without Langevin algorithm) MD run was
> also performed. What I got was a nice looking but very shallow profile in
> the case of 'normal' MD, and somewhat an anomalous profile from LD.
If by 'normal' MD you mean that you did not apply any thermostat, then it is
not a good idea. ABF is really meant for constant temperature simulations.
The problems you mention below are probably connected to insufficient
convergence. For this system, in all cases but the tightly restrained dimer
we used, long simulations (many ns) will be required to reach some kind of
convergence. We commented on that in the 2004 paper.
Best regards,
Jerome
> 2.
> Switching a part of harmonic restraints off didn't affect the depth of the
> minima; only the asymptotic value (for large Xi) would now tend to zero.
>
> 3.
> Four 0.5-ns simulations (Xi= 4.0-->6.0; 6.0-->8.0; 8.0-->10.0; 10.0-->12.0)
> produced a discontinuous deltaG dependence which tends to zero for
> Xi=12.0A.
>
> From this data, estimated deltaG of dimer dissociation is about +3
> kcal/mol.
>
> 4.
> The important point is that I wasn't able to make the program decrease
> the value of Xi. In this case, only zero deltaG values were printed to
> the output file. I could only obtain non-zero result when increasing the
> value of Xi (e.g. from 4.0 to 12.0 A), therefore simulating dissociation
> rather than association of the complex.
>
> Comments are welcome. Below: the ABF section of the input file.
> Regards,
> Grzegorz Jezierski
>
>
>
> =======================================
> abf coordinate distance-com
> abf abf1 { 11 12 16 17 }
> abf abf2 { 1 2 5 8 }
> abf xiMin 4.0
> abf xiMax 12.0
> abf dxi 0.1
> abf fullSamples 500
> abf outfile ace_guanid_dGfinal.01.dat
> abf outputFreq 100
> abf historyFile abf_forces.01.dat
> abf inFiles { }
> abf distFile abf_distrib.01.dat
> abf WriteXiFreq 100
> abf ForceConst 10.0
> abf dSmooth 0.4
>
> abf RestraintList {
> CNN_COO_plane1 { dihe {GUAN 2 H21} {GUAN 2 N2} {ACET 1 C2} {ACET 1 O1} 5.0
> 0.0 } CNN_COO_plane2 { dihe {GUAN 2 H22} {GUAN 2 N2} {ACET 1 C2} {ACET 1
> O2} 5.0 0.0 } }
This archive was generated by hypermail 2.1.6 : Wed Feb 29 2012 - 15:44:10 CST