Look the Other Way: Designing 'Positive' Molecules with Negative Data via Task Arithmetic

TL;DR

This paper introduces molecular task arithmetic, a novel approach that leverages negative data to generate positive molecules, improving diversity and success rates in molecular design tasks without needing positive examples.

Contribution

The paper presents a new method called molecular task arithmetic that uses negative data to guide the generation of positive molecules, enhancing diversity and effectiveness in molecular design.

Findings

Generated more diverse molecules than positive-data-trained models.

Successfully applied to dual-objective and few-shot design tasks.

Maintained complex properties like good docking scores.

Abstract

The scarcity of molecules with desirable properties (i.e., `positive' molecules) is an inherent bottleneck for generative molecule design. To sidestep such obstacle, here we propose molecular task arithmetic: training a model on diverse and abundant negative examples to learn 'property directions' - without accessing any positively labeled data - and moving models in the opposite property directions to generate positive molecules. When analyzed on 33 design experiments with distinct molecular entities (small molecules, proteins), model architectures, and scales, molecular task arithmetic generated more diverse and successful designs than models trained on positive molecules in general. Moreover, we employed molecular task arithmetic in dual-objective and few-shot design tasks. We find that molecular task arithmetic can consistently increase the diversity of designs while maintaining desirable complex design properties, such as good docking scores to a protein. With its simplicity, data efficiency, and performance, molecular task arithmetic bears the potential to become the de facto transfer learning strategy for de novo molecule design.