Generative design with bayesian optimization
Given a dataset of molecules with a specific property, we can employ bayesian optimization to generate molecules optimized for that property. An encoder/decoder NN can provide the ability to translate between molecules and their real vector representations. I used cddd by the Bayer group https://github.com/jrwnter/cddd but other alternatives can be plugged in.
Use the conda.txt to build the conda environment. cddd can be installed by cloning the git repository from https://github.com/jrwnter/cddd and installing with pip e.g. pip intall -e .
A notebook example is provided
cddd publication: R. Winter, F. Montanari, F. Noe and D. Clevert, Chem. Sci, 2019, https://pubs.rsc.org/en/content/articlelanding/2019/sc/c8sc04175j#!divAbstract