In this episode, we spoke with Professor Graeme Day at the University of Southampton about molecular properties prediction, especially directed towards crystal structure prediction. We discuss why a stable crystal structure is important; how machine learning can be used to help accelerate our work; and the use of blind trials to determine how good our models are.
The figure has been created using the structures obtained in Ref [1] for the, T2 (δ, β, α)polymorphs. (This includes the lowest density porous material found to this date T2-δ)
The main papers we discuss on this episode can be found in:[1] https://www.nature.com/articles/nature21419.epdf?author_access_token=MJeEVrDDiH1GTpASp-wSv9RgN0jAjWel9jnR3ZoTv0NE0WdRyYZRaYhUjIxEeWTLWj5o6FljWKmU6N3eifm4_Bz6FXdn3McgrugL2qnErff9qPrD_gGkw4sT8JB81nMC[2] https://pubs.acs.org/doi/abs/10.1021/acs.chemmater.8b01621
Furthermore, if you are curious about the machine learning methods mentioned during the episode, more information can be found in:[3] https://pubs.acs.org/doi/10.1021/acs.jctc.9b00038[4] https://pubs.rsc.org/en/content/articlelanding/2017/sc/c7sc04665k#!divAnd the blind-tests can be found at the CCDC page:[5] https://www.ccdc.cam.ac.uk/Community/Initiatives/CSPBlindTests/Credits: - Support: TMCS, EPSRC- Theme music: from "Earnest's Understanding" (2017, Charles Ormrod)
Podchaser is the ultimate destination for podcast data, search, and discovery. Learn More