Lo-Hi: Practical ML Drug Discovery Benchmark

Simon Steshin

Keywords: Molecular Property Prediction, Benchmark, Hit Identification, Lead Optimization, ML Drug Discovery

TL;DR: In Hit Identification, practitioners search for new molecules, but test ML models on molecules similar to the training set. In Molecular Optimization, researchers assume models can distinguish minor modifications, but do not test this assumption.

Abstract:

Finding new drugs is getting harder and harder. One of the hopes of drug discovery is to use machine learning models to predict molecular properties. That is why models for molecular property prediction are being developed and tested on benchmarks such as MoleculeNet. However, existing benchmarks are unrealistic and are too different from applying the models in practice. We have created a new practical \emph{Lo-Hi} benchmark consisting of two tasks: Lead Optimization (Lo) and Hit Identification (Hi), corresponding to the real drug discovery process. For the Hi task, we designed a novel molecular splitting algorithm that solves the Balanced Vertex Minimum $k$-Cut problem. We tested state-of-the-art and classic ML models, revealing which works better under practical settings. We analyzed modern benchmarks and showed that they are unrealistic and overoptimistic.

Review: https://openreview.net/forum?id=H2Yb28qGLV

Lo-Hi benchmark: https://github.com/SteshinSS/lohi_neurips2023

Lo-Hi splitter library: https://github.com/SteshinSS/lohi_splitter

Submission Number: 459