Molecule/Topology objects
| Extending the Topology API to handle biopolymers JW – Automated residue perception? JW – Built-in “unique molecules” for amino acids JC – would like to replace use of MDA, OpenMM, MDTraj Topology classes OpenMM concepts MDTraj Topology is somewhat more pythonic than OpenMM, but not perfect. But is missing chain IDs. Would like DSL (especially like Pymol DSL) Bridges with MDAnalysis – Reach out to beckstein, gowers, maybe via dotson. Functionality: Need to be able to retrieve molecule for each residue (with options for capped or uncapped, which groups to use for capping) functionality to correctly assign atom mapping for mutations Modify topology in various wayss
JC – Should have “dirty” and “clean” objects, and validate them at the time they’re used, so that users can mess around with them until that point.
Serialization for Molecule and Topology If a molecule is, at its core, a graph with nodes labeled (element, formal charge, stereochemistry, index), and edges labeled (bond order, stereochemistry), then this is easy Need to check with a cheminformatics wonk about if same graph molecule will always be interpreted the same way by aromaticity models. Some Ant’s rants imply that this is safe, basically “a given molecule with aromaticity percieved, it can be equal to many kekule structure. And all of those kekule structures WILL yield the the same aromaticity-molecule”
Need to contact RDKit and OpenEye with a test case for protein SMARTS assignment. JW will make this test case JC will contact OE/RDKit with this request. Multithreading? Just a few SMARTS per process?
Whether we should sync up with anyone on this, such as MDAnalysis: https://www.mdanalysis.org/2020/08/03/roadmap/ Slots – Probably not worth the time until parameter assignment gets really nasty Topology modification/ return_topology kwarg – Probably a good idea System should own a Topology, though it may not allow public modification.
|