Averaging over molecules of the same type

[skfegan]

We would like the entropy values to be averaged over molecules of the same type.

The difficulties are (1) defining "same type" and (2) automating the finding of the types.

(1) "same type" molecules should be of the same chemistry - for example it is obvious that water and benzene are two different types. But the environment could also play a role, for example water in the bulk could be a different type to water next to a protein. There is the question of how detailed you consider the environment (i.e. are all waters next to proteins considered the same, or are there different types for waters next to different amino acid residues)

(2) Classifying and searching for molecules of the same type can be computationally expensive (neighbour lists, etc) and clearly identifying the different environments is easier in chemical intuition than code.

[skfegan]

Averaging over the molecules of the same type makes sense when they are effectively interchangeable. Averaging at the stage of creating the force/torque covariance matrices will converge better than averaging after the matrix diagonalisation and involve fewer eigenvalue calculations than averaging entropies.

[skfegan]

If you want to group molecules to average by chemical type and environment rather than just by type, you will need to assign molecules to their type group in each frame and then collect and save the covariance matrices for each type from each frame before calculating the entropy at the end of the two loops (looping over frames and looping over molecules). Averaging by name (i.e. all water, all sodium, ...) without reference to environment would allow you to loop over the molecules and then for each molecule type loop over frames to make the matrices and calculate the entropy before moving on to the next molecule type.

[ioanaapapa]

Option to choose not to average over all molecules by choosing what to average over by index or implementing a way to identify by environment - e.g. a tetramer where only one of the monomers has a ligand bound and we want that entropy individually and average over the other three