Visualization of merged molecule.

[kexul]

Is your feature request related to a problem? Please describe.
I'm always frustrated whether the molecule generated by BioSimSpace.Align.merge is correct. That is:
1. Is each atom mapping correctly?
2. Which atoms were perturbated? Were they perturbated as expected?

Describe the solution you'd like
I hope there will be a function that generate an image or an interactive jupyter widget where the initial state and end state are showed together, with the atoms changed during the perturbation highlighted.

Describe alternatives you've considered
Here is my current solution: I saved the merged molecule to gro and top file. Open the gro file by pymol, label atoms by atom name . Then check the top file, find out which atoms would transform during perturbation.
In the following case, the place pointed by the red arrow should have an C atom appear during the perturbation. In fact, all the benzene ring should appear during the perturbation.

[lohedges]

Thanks for your question. Yes, I agree that it would be great to be able to easily visualise the merged molecule since it would help the user quickly spot potential issues prior to simulation. I had been thinking about using some of the rdkit.ChemDraw functionality, like in some of the examples from their cookbook here. Essentially I was thinking of drawing side-by-side depictions of each molecule with the MCS highlighted. I don't think it will be too hard so I'll add it to my to-do list.

Cheers.

[jmichel80]

@JenkeScheen has written some code to do so as part of his current research. See e.g.

https://github.com/michellab/dd_networkgen/blob/master/bss_mapping_checks.ipynb

[JenkeScheen]

I think this issue might interplay well with #190, potentially. If we're able to convert a (merged) BSS mol to rdkit it should be easy to visualise using standard cookbook code. As a bonus I think RDKit uses dummy atoms as well (although for different purposes); these are denoted as "*" and have atom type number 0 (1 is H, 2 is C, etc)

[kexul]

Hi @jmichel80 , thanks for the link, it quite interesting to show the mapping by atom number.

I've come up with similar idea with rdkit.Chem.Draw, but with atom color to show the atom pairs. Here it is:

I've seen some posters elsewhere, I think this one best describes what's in my mind.

Thanks for adding this to your to-do list @lohedges !

[JenkeScheen]

I vaguely remember @ppxasjsm having generated such a mapping figure before, the one with the mapping wires. Or perhaps she just showed me one that she found online, in that case sorry for @-ing you..

[lohedges]

I've made a first pass at this using some ideas from the RDkit cookbook. You can now do the following:

At present I've made it so you can look at the molecules individually, but it could be combined into a single image with a bit more work.

[jmichel80]

Neat ! I wonder if this could make its way eventually in the FEP pipeline @JenkeScheen is writing. At the moment we generate mappings automatically on a cluster after equilibration. It could make sense to generate those per ligand using non-equilibrated inputs, therefore giving the chance to a user to interactivelly adjust mappings before running a bunch of simulations downstream. If we have generated the mappings once, there should be a way to save them and re-use them once parameterised/equilibrated molecules are available.

Might be a bit tight to implement this for the coming workshop, but could be done as refinement afterwards.

[kexul]

Thank for your works! It's very impressive!

After some digging, I found another nice tool to visualize molecule xyz2graph, it creates an interactive widget to show the molecule.

I've forked it https://github.com/kexul/xyz2graph and created a molecule pair view with some modifications, here it is:

https://nbviewer.jupyter.org/github/kexul/xyz2graph/blob/33894c15ff6d0e039febc68faf53f067bcce3599/show_mol.ipynb

[lohedges]

Thanks, that looks like a nice tool. While we don't support xyz at present the format is so basic that it would be simple to write some temporary files for the purposes of making the plot. It's a shame that this package isn't available via conda-forge as we won't be able to support it. (It's dependencies are, though.)

[JenkeScheen]

So I finally found out where I'd seen that mapping method before - it's implemented in Flare! When selecting an edge in a perturbation network you can view a 3D rotatable object with the mapping:

[lohedges]

Haha. Closer to home than we thought!