CB – Words like “Topology” and “residue” have different meanings to different people. We should make sure we harmonize terminology moving forward.
CB – The connection point chemistry might be best handled by SMIRKS
CB – The amide bond parameters will be influenced by aromaticity. That may throw a wrench into capping since that may mess up linker electronic structure.
LW – Our fragmentation would keep aromatic rings together. I think that should handle this case.
CB – How do you fragment molecules into chemistries that are self-contained?
LW – This code doesn’t focus on that yet.
CB – Bill Swope would be good to involve in polymer stuff
LW – Parsnip currently is based on ecosystem-specific parameterized systems – Gromacs in this case, and handling the process of breaking and joining the whole molecule
CBy – I wonder how much of this process is Gromacs-specific? If we made this more ecosystem-agnostic, would the overall workflow would change much? There are several ways to generate coordinates, but I don’t see a way around needing to recognize residues. I think residue perception may be harder, so it may be good to focus on answers to the latter.
LW – One of my big constraints was having an object model that could help us break the polymer down into small pieces suitable for partial charge calculation and combination.
DM – There should be some advantage to getting away from atom types here, especially WRT residue perception.
CBy – How unique will “instances of the same monomer” be? If they have different neighbors.
TG – When parsnip splices things together, where are the 3d coordinates coming from?
LW – RDKit’s alignment, followed by a short minimization.
JW – I wonder if we should keep the door open to treating residues NOT JUST according to their own graph, but also that of their neighbors. OFF can already handle this with large SMARTS with some non-tagged atoms. But should that be included in the perception of the residue?
DM – How hard is bond order/formal charge perception?
CBy – This is a very hard problem. OpenEye considers this to be “toxic”. I don’t think this would be a profitable direction for LW to work on.
JW – XYZ2MOL project aims to do this. I’ve heard that they have made a very good effort at it, but it’s a very hard problem.
LW – I can’t speak for all of MDAnalysis on mol2 reading, and there are different camps. Some people are OK to support risky user behavior
DM – I advise against trying to interpret bond orders+formal charges from coordinates+elements.
Add Comment