From: Giacomo Fiorin (giacomo.fiorin_at_gmail.com)
Date: Thu Dec 10 2020 - 13:36:44 CST
Hi Zeynep, things look okay overall, so I assume for simplicity that the
computation is going as normal from the technical standpoint. Your
question is now strictly about the science itself.
As I mentioned earlier, well-tempered metadynamics is *designed to converge
numerically within a limited simulation time*. Whether this means
also a *statistical
*convergence of the PMF it is difficult to say. You need to look at the
trajectories in detail, and this cannot be done over the mailing list. If
you want to be sure, try using standard metadynamics or other methods
altogether for validation.
It is fine to apply restraints to confine sampling, e.g. by putting up wall
restraints:
https://colvars.github.io/colvars-refman-namd/colvars-refman-namd.html#sec:colvarbias_harmonic_walls
For problems like these (a very well defined geometric coordinate where it
is easy to generate multiple initial conditions) people have also used
methods that are amenable to splitting up the computation in multiple
windows to be combined later. This is not possible with metadynamics, but
is straightforward with umbrella sampling and WHAM/M-BAR, or with ABF and
the inputPrefix keyword:
https://colvars.github.io/colvars-refman-namd/colvars-refman-namd.html#sec:colvarbias_abf_params
Check out e.g. some of Jeff Comer's papers (also a contributor of this
mailing list), who happens to have used ABF on nanotubes as well. In a
different case (permeation through a quasi-solid lipid layer), I also found
it easier to generate initial structures for multiple windows, and sample
these windows in parallel.
There are also many other options that you can use, but you need to find
the tradeoff between the computer time and your own time. It is all but
guaranteed that to save one, you need to invest more of the other. I.e. we
can give you some pointers on the mailing list, but data analysis and
literature search must happen on your end.
Giacomo
On Thu, Dec 10, 2020 at 1:55 PM zeynab hosseini <hosseinizeynab93_at_gmail.com>
wrote:
> Dear All,
>
> I have performed 500ns of well-tempered metadynamics (wt-mtd) simulation
> to calculate the potential of mean force (PMF) over the adsorption of a
> small molecule on an infinite cabon nanotube surface (NAMD 2.13b1). The CNT
> is oriented along the z-axis and the reaction coordinate is the radial
> distance of molecule from the CNT central axis. The lowerboundary and
> upperboundary are considered to be 4A and 30.5A, respectively. I have
> plotted the PMF versus radial distance over 500ns (black curve), 496ns
> (red), 492ns (green), 488ns (blue) and 484ns (yellow) of wt-mtd simulation (
> link
> <https://drive.google.com/file/d/1qaJmUTdVBsy6iEZoXTcR0W7Aj9u_uVp5/view?usp=sharing>).
> The Jacobian term of the PMF (-kB T ln(r)) is included in the plot, too.
>
> I wonder if I can conclude the PMF is converged after 500ns. After about
> 450ns, the PMF plot does not show any significant change from the plot I've
> already attached. If so, then is that correct to calculate the average of
> the five mentioned PMF curves as the final PMF, along with the standard
> deviation to be considered as the error bar?
>
> Is there any way to get more accurate PMF, e.g. all five curves overlap
> each other entirley? Should I change the reaction coordinate or is there
> any other way?
>
> Is there anyway to get faster converngence? e.g. much shorter than 500ns.
> For example to reduce the upperboundary from 30.5A to 20A? I'm not sure if
> it affect the convergence positively.
>
> Since the CNT and water box are symmetrical (CNT+molecule are solvated in
> a water box of 70A × 70A × 24A), I wonder if I can confine the molecule
> movements into a specific region, in order to get faster convergence. I
> mean, e.g., if the CNT length is 20A (-10A < z < 10A), we apply restraints
> to limit the molecule movements only in a region with -5A < z < 5A ? Does
> it help to converge faster without affecting the physics of the system?
>
>
> I would really appreciate if anybody guides me..
>
> All the Best,
> Zeynab
>
> The relevant part of the metadynamics configuration file:
> colvarsTrajFrequency 500
> colvarsRestartFrequency 1000
>
> colvar {
> name Translocation
>
> upperboundary 30.5
> lowerboundary 4.0
> upperWallConstant 100.0
> lowerWallConstant 100.0
> width 0.125331414
>
> distanceXY {
> main {
> atomnumbers { 165 }
> }
> ref {
> atomnumbers {
> 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28
> 29 30 31 32 33 34 35 36 37 38 39 40 41 42 43 44 45 46 47 48 49 50 51 52 53
> 54 55 56 57 58 59 60 61 62 63 64 65 66 67 68 69 70 71 72 73 74 75 76 77 78
> 79 80 81 82 83 84 85 86 87 88 89 90 91 92 93 94 95 96 97 98 99 100 101 102
> 103 104 105 106 107 108 109 110 111 112 113 114 115 116 117 118 119 120 121
> 122 123 124 125 126 127 128 129 130 131 132 133 134 135 136 137 138 139 140
> 141 142 143 144 145 146 147 148 149 150 151 152 153 154 155 156 157 158 159
> 160
> }
> }
> axis ( 0.0, 0.0, 1.0 )
> }
> }
>
>
> metadynamics {
> name meta-distance
> colvars Translocation
> hillWeight 0.01
> newHillFrequency 250
> dumpFreeEnergyFile yes
> writeHillsTrajectory on
> hillwidth 1.253314137
> wellTempered on
> # bias factor=6
> biasTemperature 1550
>
>
> }
>
>
>
This archive was generated by hypermail 2.1.6 : Thu Dec 31 2020 - 23:17:15 CST