Technical report: Improving the properties of molecules generated by LIMO

TL;DR

This technical report explores enhancements to the LIMO framework for molecule generation, demonstrating that an autoregressive Transformer decoder with GroupSELFIES improves molecular property outcomes.

Contribution

It introduces optimized variants of LIMO, particularly using an autoregressive Transformer with GroupSELFIES, to enhance generated molecule properties.

Findings

Autoregressive Transformer with GroupSELFIES yields better molecular properties.

Ablative studies identify key factors influencing molecule quality.

Transformer-based models outperform other decoders in this framework.

Abstract

This technical report investigates variants of the Latent Inceptionism on Molecules (LIMO) framework to improve the properties of generated molecules. We conduct ablative studies of molecular representation, decoder model, and surrogate model training scheme. The experiments suggest that an autogressive Transformer decoder with GroupSELFIES achieves the best average properties for the random generation task.