I just finished reading this paper MoFlow: An Invertible Flow Model for Generating Molecular Graphs.

The paper, which is about generating molecular graphs with certain chemical properties improved the SOTA at the time of writing by a bit and used a new method to discover novel molecules. The premise of this research is that this can be used in drug design.

In the paper, they beat certain benchmarks and even create a new metric to compare themselves against existing methods. However, I kept wondering if such methods were actually used in practice. The same question is valid for any comparable generative models such as GAN's, VAE's or autoregressive generative models.

So, basically, are these models used in production already? If so, do they speed up existing molecule discovery and/or discover new molecules? If not, why not? And are there any remaining bottlenecks to be solved before this can be used?

Any further information would be great!



You must log in to answer this question.