Individual updates | DN Last week I asked for help with annual report. On Friday, Katharina Meier asked for annual report. Bayer makes their budget decisions in July. I said we’d sent the report after the annual meeting (June 28). So I’ve started the report but I still have a shallow understanding of the technical stuff. JW – As we work on the annual meeting presentation, I can work with you to propagate relevant info the the annual report. DD – DN, do you think this will be enough for your needs? DN – Yes, I see the material as largely overlapping.
MT Not much to say - Covered below in JW’s update and some stuff not of general interest Made a new version (0.0.3) of amber-ff-porting. This fixes the fractional_bondorder_method on BondHandler, and removes all sterochemistry. I verified that the files are still loadable, and that T4 lysozyme is correctly processed (compared to AMBER).
DD Most of my time went toward reaching the 0.3.0 release of PLBenchmarks. Big blocking PR is #52, with protein structure prep. Did some iterations with bobym and now it’s in a nice state. So I’ve asked IA and JC to give it an independent review. This PR will knock out a lot of the remaining issues that need to be cleared for 0.3.0 Continued optimizing approach to using Max Planck Institute cluster (MPI cluster), and now we’re getting a lot of throughput. It has a lot of compute and somewhat limited memory, so this is well suited for chapin’s sets - Looks like it’s finishing ~10k optimizations per day. I saw a throughput problem over the weekend that may be due to straining QCA’s bandwidth. Working on the S3 component for F@H. Had to make decisions about what’s hosted where. I’m going to use OMSF’s account for CI and continuous deployment. Basically, I need to be able to run unit tests with a live S3 instance. This should load-balance with the hosting that JC is going for other parts of this project.
JW – Merged PDB loading with RDKit backend, should be much faster. I’m working on PDB loading with OE backend. Right now OE-only loading won’t work, you should use an older commit if you need that. Reviewed Lorenzo’s small molecule benchmarking manuscript. It looks pretty good, you can check it out on #benchmarks-partners. Following up with MT about interchange regression tests. Partial charge changes seem limited to a small number of molecules, and the differences in assignment are also small. I think this is unlikely to be a regression introduced by interchange, the release is already months delayed, and debugging will get harder since OE code i closed source, so we’re going to put this on the backburner and move forward with Interchange. Worked with DN and LW last week to start defining our strategy/direction. Having a clear map will help us justify our value to pharma partners. Worked with MShirts and CDavel on defining CD’s summer project. Difficult because this is a chicken-and-egg problem where we’re also finding deficiencies in using tagged SMARTS as our substructure specification and networkx for graph matching.
|