Molecule optimization with MolMIM and WandB

What is MolMIM?

MolMIM is a latent variable model developed by NVIDIA that is trained in an unsupervised manner over a large-scale dataset of molecules in the form of SMILES strings. MolMIM utilizes transformer architecture to learn an informative fixed-size latent space using Mutual Information Machine (MIM) learning. MIM is a learning framework for a latent variable model which promotes informative and clustered latent codes. MolMIM can be used for sampling novel molecules from the model’s latent space. MolMIM

- Allows users to generate molecules similar to the seed molecule in SMILES format by randomly perturbing (eg by adding 0 centered gaussian noise with a desired variance) the latent space encoded from a seed molecule and decoding that back into SMILES.

- performs optimization with the CMA-ES algorithm in the model’s latent space and sample molecules with improved values of the desired scoring function. (from BioNeMo official documentation : https://docs.nvidia.com/bionemo-framework/latest/models/molmim.html)

We use MIM to avoid the main caveat of VAE, a phenomenon called posterior collaps where the learned encoding distribution closely matches the prior, and the latent codes carry little information (Razavi et al., 2019).
MolMIMは初めてMIMを分子データに適用した例です。

How to optimize molecule

Problem setting

Conclusion