Updates on projects | publication with Exscientia on FreeSolv HFEs, using different methods (e.g. OpenFF, CGenFF) – basically testing if we can do corrections with NN potentials modelforge 5 NN potentials implemented, in the process of training them Now at stage where we add docs, tutorials, so people can use it Still doing some re-factoring Close to finishing MVP phase. Not planning to add new NN potentials in near future. Projecting end of April for 1.0 release
chiron deploying trained models? collaboration with Exscientia? still in progress, setting up a marketplace for pre-trained models currently already training NN potentials for this marketplace – setting up reasonable set of hyperparameters to search over. Project that we will have trained models in the next month and the infrastructure to distribute them to people. Outward facing S3 bucket
Any further datasets work, e.g. SPICE2, metalloproteins SPICE2 PE has finished SPICE2 calculations with SPICE level of theory CI and MW getting OpenFF level of theory working Investigating why SPICE doesn’t work for some people when they want to use it for retraining Working on understanding if there is a problem with the SPICE dataset – still an open scientific question Will focus on this in the next weeks – working towards SPICE3 Focusing on the conformations generated and used
Metalloproteins
|