elixir_cz vscht


3.2Reading HILLS files
3.3Calculation of free energy surface
3.4Plotting of free energy surface
3.5Identification of minima
3.6Calculation of free energy profiles
3.7Removing one collective variable
3.8Nudged Elastic Band
4Tips and tricks
7Terms of use


Metadynminer and Metadynminer3D paper out in R Journal:
D. Trapl, V. Spiwok: Analysis of the Results of Metadynamics Simulations by metadynminer and metadynminer3d. R Journal 14(3) 46-58 (2022).

Videos from Metadynminer Plumed Masterclass (February 2022) are available at Youtube (thanks to Plumed!!!):

Video from Metadynminer webinar (31 May 2019) is available at Youtube (thanks to ELIXIR Slovenia!!!):

Metadynminer3d was published.

See comparison of accuracy of fes and fes2 vs. Plumed.


metadynminer is R packages for reading, analysis and visualization of metadynamics HILLS files produced by Plumed [1]. It reads HILLS files from Plumed, calculates free energy surfaces by fast Bias Sum algorithm [2], finds minima and analyses transition paths by Nudged Elastic Band method [3].


github https://github.com/spiwokv/metadynminer
GHactions https://github.com/spiwokv/metadynminer/actions/
appveyor https://ci.appveyor.com/project/spiwokv/metadynminer
CRAN version https://cran.r-project.org/package=metadynminer
CRAN downloads
Codecov https://codecov.io/gh/spiwokv/metadynminer/



To install R go to https://www.r-project.org/. It is available for a variety of UNIX platforms, Windows and MacOS, as well as for cloud environment. You can also use some fancy environments such as https://www.rstudio.com/.

Command lines in R enviroment start with >. Unfinished commands such as loops start with +. You can copy commands bellow (> and + are not selected) and paste them to R to run them. Output is printed in green in this document.

Install a stable version of metadynminer from CRAN by starting the R environment and typing:

> install.packages("metadynminer")
If you want to install a development version of the package from github, you need to install devtools:
> install.packages("devtools")
> library(devtools)
> devtools::install_github("spiwokv/metadynminer")

Next, invoke metadynminer:

> library(metadynminer)
Help can be obtained by the function help() with the name of the function as the argument, or by a question mark followed by the name of the function:
> help(read.hills)
> ?read.hills

Reading HILLS files

A Plumed HILLS file can be loaded by typing:
> hillsf<-read.hills("HILLS")
2D HILLS file read
This assumes that the hills file is in the working directory. If not you can specify the path (e.g. "../mymetadynamics/HILLS"). Alternatively you can check path by typing:
> getwd()
[1] "/home/yourlogin/somedirectory"
and you can change working directory by typing:
> setwd("/home/yourlogin/directory")
(may be useful in MS Windows). Hills file can be also loaded from a web site by using its URL:
> hillsf<-read.hills("http://www.metadynamics.cz/metadynminer/data/HILLS2d")
2D HILLS file read

metadynminer accepts only hills files with one or two collective variables.

If some of collective variables is periodic (usually a torsion) it is necessary to specify this fact at the moment of loading of hills file by parameter per. It must be set to a vector with one or two logical elements for one or two collective variables, respectively. In R, a vector can be created using function c() by specify its elements, e.g. per=c(TRUE, TRUE) provided that both collective variable are periodic.

> hillsf<-read.hills("HILLS", per=c(TRUE, TRUE))
2D HILLS file read
Collective variables are not periodic by default.

Period of a periodic collective variable is by default set to -π to +π. It can be changed by setting parameters pcv1 and pcv2, for example

> hillsf<-read.hills("HILLS", per=c(TRUE, TRUE), pcv1=c(0,2*pi))
2D HILLS file read
metadynminer contains two preloaded data set (two hills files), one for Alanine Dipeptide in water with Ramachandran angles φ and ψ as collective variables (called acealanme) and one for the same system with only one collective variable φ (called acealanme1d).

Metadynamics simulations are very often split into multiple runs with RESTART Plumed keyword. Hills file obtained from restarted simulations usually contain discontinuous values in the time column (i.e. they start from zero at every restart). This may complicate visualization. It is possible to solve this problem when loading hills by setting ignoretime=TRUE in the read.hills function. It will take the time of addition of the first hill and use multiplicates of this value for all other hills.

However, a user can load hills from restarted simulations without this option. First, wrong time values do not affect calculations of free energy surfaces. Second, the option ignoretime=TRUE can be used later for selected functions (see later).

The package metadynminer provides print and summary functions for its objects, including hills file. Try it by typing:

> acealanme
2D hills file HILLS2d with 30000 lines
> summary(acealanme)
2D hills file HILLS2d with 30000 lines
The CV1 ranges from -3.141309 to 3.14159
The CV2 ranges from -3.141321 to 3.141371
Hills file objects can be concatenated by addition (+) operation:
> acealanme+acealanme
2D hills file HILLS2d with 60000 lines
metadynminer provides plot function for its objects. Hills file with one collective variable will be pllotted as the evolution of the collective variable.
> plot(acealanme1d)
hills 1D

Hills file with two collective variables will be plotted as collective variable 1 vs. collective variable 2.

> plot(acealanme)
hills 2D

In a plot you can modify labels of axes (xlab and ylab), title (main) and subtitle (sub). You can also modify the range of axes (xlim and ylim). It is alo possible to modify shape, color, background and size of points, line width and plot size ratio by pch, col, bg, cex, lwd and asp, respectively (see R functions plot, points and lines).

It is possible to use ignoretime=TRUE for this function (for hills with one collective variable) if you did not used this option in read.hills.

Beside the function plot it is possible to use functions points and lines for a hills file object to plot multiple variables onto a single canvas (see R functions points and lines).

In well tempered metadynamics [4] it is useful to see evolution of heights of hills. This can be ploted by typing:

> plotheights(acealanme)

It is again possible to use ignoretime=TRUE for this function if you have not used this option in read.hills.

Calculation of free energy surface

Free energy surface is calculated by summing hills using function fes:
> tfes<-fes(hillsf)
Summing hills is much less expensive than a metadynamics simulation, nevertheless, it can be a slow process if number of hills exceeds 10,000. We developed a Bias Sum algorithm [2] to accelerate this calculation. Instead of evaluation exp(-||s-St||2/2σ2) for every hill it calculates a single Gaussian hill and relocates it. This function is not available for simulations with variable hill widths. As an alternative it is possible to calculate free energy surface by conventional way (function fes2). This function works with variable hill widths:
> tfes<-fes2(hillsf)
Hills can be summed for whole simulation (default) or for selected hills. The range can be controlled by imin and imax. These are indexes of first and last hill to be summed.
> tfes<-fes(hillsf, imin=5000, imax=10000)
By default the free energy surface is calculated for 256 (1D) or 256x256 (2D) grid points. This can be changed by npoints option.

Two free energy surfaces can be added and subtracted:

> tfes
2D free energy surface with 256 x 256 points, maximum -23.96706 and minimum -97.32465
> tfes+tfes
2D free energy surface with 256 x 256 points, maximum -47.93411 and minimum -194.6493
> tfes-tfes
Warning: FES obtained by subtraction of two FESes
 will inherit hills only from the first FES
2D free energy surface with 256 x 256 points, maximum 0 and minimum 0
It is possible to add or subtract a constant. Free energy surface can be also multiplied or divided by a constant:
> tfes/4.184
Warning: division of FES will divide
 the FES but not hill heights
2D free energy surface with 256 x 256 points, maximum -5.728264 and minimum -23.26115
It is possible to calculate minimum and maximum of free energy surface by functions min and max, respectively. This can be used, for example, to convert a free energy surface so that the global minimum is equal to zero:
> tfes-min(tfes)
2D free energy surface with 256 x 256 points, maximum 73.35759 and minimum 0
The function summary function gives the same output as print for free energy surface object:
2D free energy surface with 256 x 256 points, maximum -23.96706 and minimum -97.32465
2D free energy surface with 256 x 256 points, maximum -23.96706 and minimum -97.32465

Plotting of free energy surface

Free energy surface can be plotted by function plot. This function behaves differently depending on dimensionality of the free energy surface. For 1D surface it plots the profile:
> tfes<-fes(acealanme1d)
> plot(tfes)
free energy surface 1D

For 2D surface it plots the free energy surface by R function image, contours or by combination of both (default):

> tfes<-fes(acealanme)
> plot(tfes)
free energy surface 2D

The type of a 2D plot can be controlled by plottype option (can be set to "image", "contour" or "both").

Similarly to hills file you can modify labels of axes (xlab and ylab), title (main) and subtitle (sub) and range of axes (xlim, ylim and zlim). If you exceed the range of calculated free energy surface it will plot blank areas. To avoid this you can specify xlim and/or ylim in the fes function.

For 1D surface you can modify the color of line by col option (see colors to see colors available in R).

For 2D surface it is possible to chose whether it is ploted as a heatmap (R function image), contours (R function contours) or by combination of both (default). For image you can modify the color scale (default is rainbow(135)[100:1]). See help(rainbow) to see color palettes in R. For contour plot you can change contour values by nlevels. For example to have free energy range 0-200 kJ/mol with 20 contours you can use:

> tfes<-fes(acealanme)
> tfes<-tfes-min(tfes)
> plot(tfes, zlim=c(0,200), nlevels=20)
free energy surface 2D

You can also implicitly specify contour values explicitely by levels:

> plot(tfes, zlim=c(0,200), levels=1:20*10)
Here 1:20 returns 1, 2 ... 20 and 1:20*10 return 10, 20 ... 200.

Color scale can be printed by option colscale=TRUE:

> tfes<-fes(acealanme)
> plot(tfes, colscale=TRUE)
free energy surface 2D

Identification of minima

Free energy minima corresponds to stable or metastable states of the studied system. You can use function fesminima to automatically locate free energy minima on the free energy surface:
> tfes<-fes(acealanme)
> minima<-fesminima(tfes)
This function divides the free energy surface into 8 (1D) or 8x8 (2D) equivalent bins. For each bin it finds its energy minimum. Next, it checks whether these minima are local minima of the whole surface. These minima are collected, sorted and labeled by alphabet letters. The number of bins can be modified by parameter nbins.

Functions print and summary are available for a minima object:

> minima

  letter CV1bin CV2bin        CV1        CV2 free_energy
1      A     78    237 -1.2443171  2.6734337   -97.32465
2      B     29    241 -2.4516743  2.7719935   -95.77144
3      C     75    118 -1.3182369 -0.2587194   -94.66432
4      D    167    152  0.9486378  0.5790386   -91.86909
5      E    170    255  1.0225576  3.1169527   -84.57032
> summary(minima)
[1] 6
  letter CV1bin CV2bin        CV1        CV2 free_energy relative_pop
1      A     78    237 -1.2443171  2.6734337   -97.32465 8.837680e+16
2      B     29    241 -2.4516743  2.7719935   -95.77144 4.741231e+16
3      C     75    118 -1.3182369 -0.2587194   -94.66432 3.041705e+16
4      D    167    152  0.9486378  0.5790386   -91.86909 9.917603e+15
5      E    170    255  1.0225576  3.1169527   -84.57032 5.315355e+14
1 50.0278223
2 26.8388849
3 17.2183058
4  5.6140985
5  0.3008885
The function summary calculate equilibrium populations of each minimum in percent. Keep in mind that these values were calculated from free energies of minima points, not by integrating populations of each free energy minimum basin. These values are dependent on temperature and energy units. By default these are set to temp=300 and eunit="kJ/mol". The parameter eunit can be set to "kJ/mol" or "kcal/mol".

Sometimes it is useful to add some point in the space of collective variables to the list of automatically identified minima. This can be done by making a "fake" minimum and adding it to the list:

> minima<-fesminima(tfes)+oneminimum(tfes, cv1=0, cv2=0)
> minima

  letter CV1bin CV2bin        CV1        CV2 free_energy
1      A     78    237 -1.2443171  2.6734337   -97.32465
2      B     29    241 -2.4516743  2.7719935   -95.77144
3      C     75    118 -1.3182369 -0.2587194   -94.66432
4      D    167    152  0.9486378  0.5790386   -91.86909
5      E    170    255  1.0225576  3.1169527   -84.57032
6      F    129    129  0.0000000  0.0000000   -49.33291
This will create the same set of minima with the additional one placed in the point {0,0}.

The function plot applied on a minima object provides a same image as for the corresponding free energy energy minimum object with additional minima labels:

> plot(minima)
minima 2D

Color of labels can be controlled by textcol option.

Calculation of free energy profiles

Imagine metadynamics of a system with two energy minima, one (A) at 0 kJ/mol and the second (B) at 10 kJ/mol, separated by a transition state (TS) at 30 kJ/mol. We want to estimate the free energy difference between B and A. At the begining of metadynamics it is estimated as zero, because there is no bias potential. Metadynamics starting from B would flood this minimum after some time. At this point the free energy estimate is -20 kJ/mol. Next, the system jumps to A. After flooding A the free energy estimate reaches the correct value of 10 kJ/mol. After this point the system can freely jump from A to B and B to A because both minima are flooded and the free energy surface is flattened. This leads to oscillation of free energy difference around the correct value of 10 kJ/mol. It is therefore useful to examine the evolution of free energy difference during the simulation. Convergence of these values indicate (but do not fully proof) that the estimated free energy difference is accurate.

This can be evaluated by function feprof:

> tfes<-fes(acealanme)
> minima<-fesminima(tfes)
> prof<-feprof(minima)
This function calculates evolution of free energy differences between each minimum and the global minimum (global at the end of simulation).Free energy of the global minimum is always zero.

Similarly to functions fes and fes2 it is possible to specify the range of hills by imax (imin option is not available because it does not make sense to calculate profiles from other point than from the start).

It is possible to use functions print and summary on free energy profile:

> prof
30001 2D minima
> summary(prof)
  letter CV1bin CV2bin        CV1        CV2 free_energy   min diff  max diff
1      A     78    237 -1.2443171  2.6734337   -97.32465  0.0000000  0.000000
2      B     29    241 -2.4516743  2.7719935   -95.77144 -2.0561128  3.731685
3      C     75    118 -1.3182369 -0.2587194   -94.66432  0.3887439 12.213005
4      D    167    152  0.9486378  0.5790386   -91.86909  0.7606614 21.746684
5      E    170    255  1.0225576  3.1169527   -84.57032  0.8638223 21.746684
1  0.000000
2  1.637328
3  2.670321
4  5.497148
5 12.875954
The function summary prints minimal and maximal free energy difference for time range specified by imind and imaxd. For example, it is possible to calculate the free energy surface from all hills and evaluate convergence of the free energy difference in last 10,000 hills:
> tfes<-fes(acealanme)
> minima<-fesminima(tfes)
> prof<-feprof(minima)
> summary(prof, imind=20000, imaxd=30000)
  letter CV1bin CV2bin        CV1        CV2 free_energy   min diff  max diff
1      A     78    237 -1.2443171  2.6734337   -97.32465  0.0000000  0.000000
2      B     29    241 -2.4516743  2.7719935   -95.77144  0.4399352  1.700877
3      C     75    118 -1.3182369 -0.2587194   -94.66432  1.1614741  3.229472
4      D    167    152  0.9486378  0.5790386   -91.86909  4.4346671  5.967618
5      E    170    255  1.0225576  3.1169527   -84.57032 11.0699007 13.173305
1  0.000000
2  1.637328
3  2.670321
4  5.497148
5 12.875954
This shows that the free energy of B (realtive to A) was 0.4399352-1.700877 and value at the end of simulation (in imaxd) was 1.637328 kJ/mol.

Profiles can be plotted using plot function:

> plot(prof)
profiles 2D

By default it plots profiles for all free energy minima. It is possible to specify minima to be plotted by which followed by the vector of indexes of minima.

Removing one collective variable

Sometimes it may be useful to get rid of one collective variable by calculating a 1D free energy surface from a 2D one. It is not possible to do this simply by ignoring on collective variable in the hills summing process. Instead, it is necessary to convert 2D free energy surface into probabilities (as exp(-F/kT)), integrate probabilities along one collective variable and convert probabilities back to free energies (as -kTlog(P)). This can be done by function fes2d21d:
> tfes1<-fes2d21d(acealanme, remdim=2)
> plot(tfes1)
free energy surface 2D to 1D

The process depends on temperature of simulation and energy units. It is therefore possible to change them by temp and eunit options. The output of fes2d21d is a 1D free energy surface. Similarly to fes it is possible to set imin, imax, xlim, xmax and npoints.

Nudged Elastic Band

One of motivation of molecular modeling is identification of transition paths and transition states. In metadynminer it is possible to use Nudged Elastic Band method [3] implemented in the function neb. First it is necessary to calculate free energy minima. By visual inspection we can decide that it would be interesting to find a transition path between minima A and D. The path can be analyzed by:
> tfes<-fes(acealanme)
> minima<-fesminima(tfes)
> myneb<-neb(minima, min1="A", min2="D")
> myneb
path between minima:
  letter CV1bin CV2bin       CV1      CV2 free_energy
1      A     80    235 -1.195037 2.624154    -96.5517
  letter CV1bin CV2bin       CV1       CV2 free_energy
4      D    167    152 0.9486378 0.5790386   -91.86909
> summary(myneb)
path between minima:
  letter CV1bin CV2bin       CV1      CV2 free_energy
1      A     80    235 -1.195037 2.624154    -96.5517
  letter CV1bin CV2bin       CV1       CV2 free_energy
4      D    167    152 0.9486378 0.5790386   -91.86909
with kinetics
  direction   deltag   halflife units
1        -> 24.07849   1.726768    ns
2        <- 19.39587 264.200779    ps
Minima min1 and min2 can be specified by letters or indexes.

The function print prints minima, the function summary predicts kinetics (half times) for A to D and D to A transitions. For this it uses Eyring equation [5] with transition coefficients set to 1. By default, Nudged Elastic Band has 20 data points (can be changed by nbins). It may happen that a path identified by neb is wrong. For example it hapens for the pair of minima C and D. Then it is necessary to modify parameters of Nudged Elastic Band, namely the number of steps (nsteps, default 100), step size (step, default 1) and toughness of the band (k, default 0.2). For the pair of minima C and D it is sufficient to set nsteps=1000 to get a realistic path.

Path identified by Nudged Elastic Band can be ploted:

> plot(myneb)
Nudged Elastic Band 1

Alternatively, it is possible to plot it projected onto a free energy surface. This can be done by this pair of commands (without closing the image window after the first command). The second function works similarly to points or lines functions:

> plot(minima)
> linesonfes(myneb)
Nudged Elastic Band 2

Tips and Tricks

Publication quality figures

Following script can be used to generate a publication quality figure (8x8 cm, 600 dpi):
> tfes<-fes(acealanme)
> png("filename.png", height=8, width=8, units='cm', res=600, pointsize=6)
> plot(tfes)
> dev.off()
You can replace png by bmp, jpeg, tiff, pdf or eps to get other file formates (parameters can be different).

Making FES relative to the global minimum

You can set free energy minimum to zero by typing:
> tfes<-tfes-min(tfes)

Hills from restarted simulations

There are two ways to cope with hills file from restarted simulations, i.e. simulations where the time column starts from zero at every restart. It is possible to set ignoretime=T in the read.hills function. This takes the time of the first hill and use it as the uniform step.

Alternatively, it is possible to keep original times and use ignoretime=T for other ploting functions.

Making movie

Individual snapshots of a movie can be generated by:
> tfes<-fes(acealanme, tmax=100)
> png("snap%04d.png")
> plot(acealanme, zlim=c(-200,0))
> for(i in 1:299) {
+   tfes<-tfes+fes(acealanme, imin=100*i+1, imax=100*(i+1))
+   plot(tfes, zlim=c(-200,0))
> }
> dev.off()
These files can be concatenated by a movie making program such as mencoder. You can modify png functions to modify size and resolution.

If you instead want to see flooding, type:

> tfes<-fes(acealanme)
> png("snap%04d.png")
> plot(tfes, zlim=c(-200,0))
> for(i in 0:299) {
+   tfes<-tfes+ -1*fes(acealanme, imin=100*i+1, imax=100*(i+1))
+   plot(tfes, zlim=c(-200,0))
> }
> dev.off()

Evaluation of convergence of one CV

You can use function fes2d21d to convert a 2D surface to 1D and to evaluate the convergence:
> tfes1<-fes2d21d(acealanme, remdim=2)
> plot(tfes1-min(tfes1), ylim=c(0,80), lwd=4, col="black")
> for(i in 1:10) {
+   tfes1<-fes2d21d(acealanme, imax=3000*i)
+   lines(tfes1-min(tfes1), col=rainbow(13)[i])
> }

Transforming CVs

If you want, for example, to use degrees instead of radians on axes, set axes=F in the plot function and then plot (without closing the plot window) both axes separately.
> plot(tfes, axes=FALSE)
> axis(2, at=-3:3*pi/3, labels=-3:3*60)
> axis(1, at=-3:3*pi/3, labels=-3:3*60)
> box()
The expression -3:3 will generate a vector {-3,-2,-1,0,1,2,3}, which can be multiplied by π/3 (tick positions in radians) or by 60 (tick positions in degrees). If you want to transform just one axis, e.g. the horizontal one, while keeping the vertical unchanged, simply type axis(2) for the vertical one. The function box() redraws the box.

kcal vs kJ

MetadynMiner works in kJ/mol by default. If your MD engine uses kcal/mol instead, you can either multiply your free energy surface by 4.184 to get kJ/mol. If you prefer to keep kcal/mol, you can set eunit="kcal/mol" for functions fes2d21d or summary of minima and neb objects. Other units are not supported.

Shifting a periodic CV

Results of some simulations with a torsion collective variable may be difficult to analyze and visualize in the range -π - +π. It may be useful to shift collective variables to 0 - 2 π. However, this problem is not very common so we did not prepare any user friendly way how to solve this. It can be solved in a user unfriendly way. Let us consider we want to shift the first collective variable to be in the range 0 - 2 π. First we will make a copy of acealanme. We will change its pcv1 to c(0,2*pi). Finally we can add 2 π to negative values of the first collective variable:
> acealanmec<-acealanme
> acealanmec$pcv1<-c(0,2*pi)
> acealanmec$hillsfile[acealanmec$hillsfile[,2]<0,2]<-
+     acealanmec$hillsfile[acealanmec$hillsfile[,2]<0,2]+2*pi
> tfes<-fes(acealanmec)
> plot(tfes)
The hills file object has several instances including hillsfile, which contains the HILLS file, and pcv1 with collective variable periodicity. They can be printed using $ operator. The expression acealanmec$hillsfile[,2] prints all values of the first collective variable (second column). The expression acealanmec$hillsfile[,2]<0 prints the same number of TRUE or FALSE values depending whether the first collective variable is positive or negative. The expression acealanmec$hillsfile[acealanmec$hillsfile[,2]<0,2] prints only negative values of the first collective variable. They can be replaced by the same value + 2 π.


  1. G.A. Tribello, M. Bonomi, D. Branduardi, C. Camilloni, G. Bussi: PLUMED2: New feathers for an old bird. Comput. Phys. Commun. 185, 604 (2014).
  2. P. Hosek, V. Spiwok: Metadyn View: Fast web-based viewer of free energy surfaces calculated by metadynamics. Comput. Phys. Commun. 198, 222 (2016).
  3. G. Henkelman, H. Jonsson: Improved tangent estimate in the nudged elastic band method for finding minimum energy paths and saddle points. J. Chem. Phys. 113, 9978 (2000).
  4. A. Barducci, G. Bussi, M. Parrinello: Well-tempered metadynamics: A smoothly converging and tunable free-energy method. Phys. Rev. Lett. 100, 020603 (2008).
  5. H. Eyring: The activated complex in chemical reactions. J. Chem. Phys 3, 107 (1935).


Author + maintenance: Vojtech Spiwok (spiwokv<you-know-what>vscht.cz)

Terms of use

Terms of use