fgrunewald/fast_forward

Map atomistic trajectories to CG Martini level fast and compute interactions

fast_forward's Issues

Output overwrites exisiting files

If the output file has the same name as an existing file, the existing file is not backed up.

Water mapping

Would be cool to also have a good procedure to map water.

Angle potential parameters with etropic correction

Currently, the program calculates estimates of the angle bending parameters directly from the distribution of the angle from AA trajectory.
This is correct for angles in restrained environment, like in sugar rings,
but is incorrect for angles of freely moving moieties as e.g. in alkanes.

In alkanes, the free movement of the chain favours entropically perpendicular angles (i.e. around 90º)
and disfavours those around 0º and 180º.
For straight chains, this has a qualitative effect – the resulting estimates are equilibrium angle, e.g. 140º instead of 180º.
See e.g. here, angle from pycgmap-ped Charmm36 (blue: current status, orange: with the entropy correction)

The correction is simple – use the Jacobi factor of the transformation to remove this entropy from the distribution.
This can be simply done by assigning weights to the angle values
w(ang) = 1.0/sin(ang)

I think this could be implemented either as an CLI option or as a comment in the input itp file.

Support for melt systems

Currently only single molecule systems are supported

Reading in mapping files vs. directory of .map files

Dihedral timeseries not computed

Resid 0

What happens when we set the resid to 0. This relates to the old PR #5 .

Raise Error if no coordinates are selected

Overlapping mappings between residues

At the moment, if an atom is mapped between two residues, it can't be accounted for in both of them.

tests needed

We need some tests to make sure core functionalities work as intended

Outdated dependencies make fast_forward difficult to install

Installing fast_forward using any version of python > 3.8.18 is impossible due to the dependency on numpy version 1.20.0.

Additionally, some package managers will try to install other dependencies which will cause another clash. For example, using conda to install fast_forward will result in conda trying install a gsd version greater than 3.2.0 for MDAnalysis. The result is a failing installation due to the numpy version.

Being limited to python 3.8.18 is also preventing the use of newer MDAnalysis versions. This means only .tpr files from Gromcas version 2021 or lower can be read.

Interaction Class iter should skip unkown interaction types

Single resname gives error

Currently using a single resname gives an error. Defaults in the map parser need fixing.

Sign problem of angles

Since 'np.arccos' is used to calculate the angle distributions, the returned value is always positive. This is the convention used by Numpy. This, however, causes some sign issues while fitting distributions.

A possible solution is calculating the determinant of 'a matrix' to get the correct sign, see here.

PBC completion super slow

There should be a faster way to do the PBC completion. Subprocess call to GROMACS?