bug-frame-charge-94

[universvm]

Closes #94 Fix

[ChrisWellsWood]

Looks good, approved. One thought on this that we should discuss though, your threshold for the Zimmerman polarity is quite high, so only a few residues are counted as polar, and this doesn't include residues like serine or threonine, which are traditionally included as such. We might want to explore using Grantham polarity with a cut off around 8 too.

[sunal1996]

In this part of create_frame_dataset.py

            gaussian_atom = gaussian_matrix[:, :, :, atom_idx]
            # Add at position:
            frame = add_gaussian_at_position(
                main_matrix=frame,
                secondary_matrix=gaussian_atom,
                atom_coord=indices,
                atom_idx=atom_idx,
            )
            if res_property != 0:
                gaussian_atom = gaussian_matrix[:, :, :, atom_idx] * float(res_property)
                # Add at position:
                frame = add_gaussian_at_position(
                    main_matrix=frame,
                    secondary_matrix=gaussian_atom,
                    atom_coord=indices,
                    atom_idx=5,
                    normalize=False,
                )
                

Does it appear that we are declaring the gaussian_atom variable and running add_gaussian_at_position function redundantly? In that case, does this one make sense:

            if res_property != 0:
               gaussian_atom = gaussian_matrix[:, :, :, atom_idx] * float(res_property)
               # Add at position:
               frame = add_gaussian_at_position(
                   main_matrix=frame,
                   secondary_matrix=gaussian_atom,
                   atom_coord=indices,
                   atom_idx=5,
                   normalize=False,
             else:
                gaussian_atom = gaussian_matrix[:, :, :, atom_idx]
                  # Add at position:
                frame = add_gaussian_at_position(
                     main_matrix=frame,
                     secondary_matrix=gaussian_atom,
                     atom_coord=indices,
                     atom_idx=atom_idx,
                    )

[universvm]

@sunal1996 in the case of a charged residue (say lysine), the gaussian needs to be added in the atom index (C, N, O, Ca, Cb) AND in the charge index (ie. the fifth).

I think in your if/else logic all the atoms of a charged residue (C, N, O, Ca, Cb) would end up in the 5th channel and there would be no atoms in the C, N, O, Ca, Cb channels

[sunal1996]

Ah, all clear.

[ChrisWellsWood]

Looks good merge when ready.