Ecosyste.ms: Awesome

An open API service indexing awesome lists of open source software.

Awesome Lists | Featured Topics | Projects

https://github.com/whitead/plumed_grids

A Python script to aid creating and manipulating the grids which are used by plumed and plumed2
https://github.com/whitead/plumed_grids

Last synced: about 1 month ago
JSON representation

A Python script to aid creating and manipulating the grids which are used by plumed and plumed2

Awesome Lists containing this project

README 

        
Plumed Grids

====



A Python script to aid creating and manipulating the grids which are

used by plumed and plumed2. Just add the location of the folder to

your PYTHONPATH with



    export PYTHONPATH=$PYTHONPATH:/path/to/this/repo



It does some nice things, but doesn't work with derivatives and does

not do spline interpolation.



Quickstart

----



```python

from plumed_grids import *



#create empty grid

g = Grid()



#read in a PLUMED grid. In this example, it's assumed to be 2D

g.read_plumed_grid('BIAS')



#Add a pretty sunset to the bias

g.add_png_to_grid('sunset.png')



#normalize it so it integrates to 1

g.normalize()

print g.integrate_all()



#plot it to make sure it looks good

g.plot_2d('my_bias.png')



#write it to be ready for plumed2

g.write_plumed2_grid('plumed_2.dat', names=['phi', 'psi', 'pot.target'])

```



The Grid Object

---



```python



g = Grid()



#Create grid by adding CVs. The name doesn't matter so much, but was tracked in PLUMED1 (not PLUMED2)

#After name comes min, max, nbins, and periodicity

g.add_cv('Absolute position', -155, 155, 1024, False)

g.add_cv('Absolute position', -155, 155, 1024, False)



#add some values to bias using an image

g.add_png_to_grid('test.png')



#see a summary 

print g



#see individual attributes

print g.nbins

print g.dims

print g.dx

print g.ncv



#Write/Read

g.write_plumed1_grid('test_1.dat')

g2 = Grid()

g2.read_plumed1_grid('test_1.dat')



#uses fuzzy floating point to test equality

#useful for testing (perhaps one day for unit tests of PLUMED/PLUMED2)

print g == g2



```



Plumed 1 vs Plumed 2

----



When writing files, you must specify with either `write_plumed1_grid`

or `write_plumed2_grid`. When reading, the generic `read_plumed_grid`

will infer the file type.



When writing Plumed 2 grids, you can optionally specify names by

passing a list to the `write_plumed2_grid` method 



Integrals and Units

---



It is always assumed that what is stored in the grid is -ln P(s) or F(s) / kT, where P(s) and F(s) are probability and free energy, respectively. To integrate we take e^(-V) of the values (V) stored and then return -log

of that integral. So, if units are in energy in kT, we integrate the

probability. This is the format used for experiment directed

metadynamics (EDM) as implemented in Plumed 2, Plumed 1, and the

electronic dance music lammps plugin.



Regions

----



```python



#Continuing the example from above:



#Create a function which returns True when in region or False

def ellipse_region(x, center, radii):

    sum = 0

     for xi, ci, ri in zip(x, center, radii):

	 sum += (xi - ci) ** 2 / ri ** 2

	      if(sum > 1):

	          return False

	  return True

				    

#create a few particular ellipses

region1 = lambda x: ellipse_region(x, [0,0], [2.5, 5])

region2 = lambda x: ellipse_region(x, [-2,4], [2.5, 5])



#plot them to make sure they're correct

g.plot_2d_region('region1.png', region1)

g.plot_2d_region('region2.png', region2)



#get free energy difference between them

-np.log(g.integrate_region(region1)) + np.log(g.integrate_region(region2))

```



Convert Plumed 1 to Plumed 2

----



Note that the derivatives will not be transferred!



```python

g = Grid()

g.read_plumed1_grid('some_grid.dat')

g.write_plumed2_grid('some_grid2.dat', names = ['foo', 'bar', 'metad.target'])

```



Manipulate Grids

----



```python

g = Grid()



#read in grid previosuly created to get its min/max/bins 

g.read_plumed1_grid('bias.dat')



#zero out its potential

g.pot[:,:] = 0



#add a picture to it

g.add_png_to_grid('logo.png')

g.normalize()



#Make it so that the maximum difference is 3 kT

diff = np.max(g.pot) - np.min(g.pot)

g.pot *= 3. / diff

g.pot -= np.max(g.pot)

g.plot_2d('logo-bias.png', axis=(0,1))

g.write_plumed1_grid(open('logo.dat', 'w'))

g.pot *= -kt

g.pot -= np.max(g.pot)

g.plot_2d('logo-fes.png',axis=(0,1))



######################



g = Grid()

g.read_plumed2_grid('bias.dat')



#Change the bin number with interpolation

g.set_bin_number( (1024, 1024) )



#Stretch by 20% 

g.stretch( (1.2, 1.2) )



#Smooth the bias by adding a blur with a radius of 3 in x and 2 in y

g.gaussian_blur( (3,2) )



#Extend the edge value outwards 20% so that there is a margin. Make it a whole number (pretty = True)

g.add_margin( (1.2, 1.2), pretty=True)



```



Requirements

---



Numpy, Scipy, Matplotlib (if plotting is used)



Disclaimer

----



Grid orientation can be a funny thing. Please make sure by examining

the min/max and number of bins that your grid is correctly oriented in

PLUMED/PLUMED2 and when read into the python code. Plotting a lot

helps too.



Cite

----



This was created for helping in the experiment directed targeted

metadynamics work. Please cite:



**Designing Free Energy Surfaces that Match Experimental Data with Metadynamics**. AD White, JF Dama, GA Voth. *J. Chem. Theory Comput.* **2015**, *11 (6)*, pp 2451-2460