Working on small refinements for perceive_residues – Can now recognize some small caps, can support using different substructure libraries.
Adding option to use strict chirality or not - Useful in cyclic peptides.
JW – Atom (@) vs. bond (/ and \), aromatic bonds (sometimes have @ in them)? I’ll add a to-do to the PR and we can inherit this into the big biopolymer PR.
Putting together from_pdb functionality for molecules. Currently early prototype/WIP.
Asked Perses team about possibly extending PDBFixer/making a onramp from PDBFixer output. They said they’ve had trouble with it, especially WRT missing atoms and stereochemistry.
JC suggested talking with Tristan Croll/Phenix devs. They’ve already solved the problem of reading from PDB.
CC
Doing lots of PI wrangling for review paper. Now have people signed up to write drafts of sections for room temp xtals, NMR.
Working on problem with generating ring conformers for proline. Would like to use this for setting up torsiondrives. OE has had a lot of trouble with doing ACE-capped N terminal prolines (OE doesn’t recognize this). Because it’s not recognized as a residue, you can’t drive the backbone/sidechain torsions, so instead I need to manually assign them. But there’s some further technical problem as well, maybe the backbone ring torsion overlaps with another ring torsion and causes problems.
IP – Espaloma people had a similar problem with OpenEye. They used the topology-refactor branch of the toolkit to deal with it – This is actually why I added caps. So this should let you identify which atoms belong to caps vs AAs.
MT
Climbed pikes peak
Gave a talk on Interchange to Chodera lab Thursday morning. Spent a good amount of the week doing prep for that. Currently it’s mostly talk since there’s some missing gaps in implementation. I had hoped to get more direct feedback/feature requests, but questions were more scientific. One good idea that came out of discussion w/ Mike Henry was to use Interchange MD engine drivers to compare different versions of same engine (as well as different engines).
JW – Was this recorded?
MT – Yes, JChodera has a copy, can ping him. Also the slides are available on internal GDrive and should be suitable for stitching into other presentations.
Worked on prototype for swapping OpenMM to OpenFF units - Can basically make OpenMM systems for standard workflows. So biggest blockers will be to decide when to make the big API break.
This week, will push to do one-on-ones with OpenFF scientists to get Interchange implemented into fitting workflows. Meeting wiht PB later today on this.
JW – Hoping to do roadmap review at next all-hands meeting. Could be a good chance to get high-level buy in on getting interchange
MT – I’d like to get this rolling on a shorter timeframe, so I’ll reach out to people directly.
PB – Will anything need to change in ForceBalance?
JW – I don’t think so – create_openmm_system will remain, just the internals will be
Will be working on design+architecture proposals. JW and I will work on desired QCArchive infrastructure+refactors - Basically articulating goals for QCA developer.
Will schedule meeting with JW on capacity/task planning this week.
This week, will be working on refactoring PLBenchmarks to run on F@H.
Benchmarking:
LD’Amore is taking over collection for Sage results from industry partners/datasets. I’m supporting him on this.
Want to send out a survey to see how people likes specific aspects of season 1, what they’d want from a season 2.
PB
Moving forward with theory benchmarking – Writing and pushing forward another dataset.
Helping rotation students with doing some experiments – Getting evaluator working, science support,
Will be pushing again on WBO stuff – CBayly wanted to take another look at this so I’ll be doing another biaryl set. I need help with setting up a local QCF instance going at UCI.
DD – I’d be happy to help, I’m free after 11 AM
(General) – PB and DD will meet later today.
JW – For WBO fitting, it may be helpful to use ELF conformer selection. I wonder if it’s posible to thread that into the WBO fitting experiments – Thsi may reduce the amount of noise from random conf gen.
PB --Would this just be using AM1BCCELF10 as the method?
JW – I think so. Happy to talk about this Weds or Thurs.
JW
Biopolymer refactor
Initial implementation of residues + other iterators
Requires deprecation of TopologyMolecules, TopologyAtoms, etc. Big API change.
Checked in with MolSSI (Andrew Abi-Mansour / Ben Pritchard) on MMIC/MMSchema.
Started updating annual roadmap w/ Matt. This was to help us get resolution of what we should eb doing on the yearish timescale, feel free to propose other big items. Infrastructure Roadmap 2022
Botched Forcebalance release, will fix today.
FB had a few imports of the form from simtk.openmm import _openmm. I wrapped this in a try/except but accidentally made the import “hard” instead of two places where it should have been “soft”. So I made a botched 1.9.1 release where openmm is a hard dep, and today or tomorrow I’ll make a 1.9.2 with this fixed.
JW – Is this worth making additional tests/refactoring FB tests for?
DD – Probably not
MT – This may be worth refactoring – It imposes unknown costs/emergencies and that adds up over a year. It depends on how far our responsibility extends, and how complex it would be to implement “bleeding edge” tests that would catch things like the openmm release/namespace change.