Modeling Industrial ADMET Data with Multitask Networks

TL;DR

This paper evaluates the effectiveness of multitask neural networks in industrial ADMET data modeling, showing modest benefits and dataset-dependent effects, with implications for optimizing drug discovery models.

Contribution

It provides a comprehensive comparison of neural networks versus baseline models on ADMET datasets and analyzes how dataset size and side information influence multitask learning performance.

Findings

Multitask learning offers modest improvements over single-task models.

Smaller datasets benefit more from multitask learning.

Adding extensive side information does not always enhance performance.

Abstract

Deep learning methods such as multitask neural networks have recently been applied to ligand-based virtual screening and other drug discovery applications. Using a set of industrial ADMET datasets, we compare neural networks to standard baseline models and analyze multitask learning effects with both random cross-validation and a more relevant temporal validation scheme. We confirm that multitask learning can provide modest benefits over single-task models and show that smaller datasets tend to benefit more than larger datasets from multitask learning. Additionally, we find that adding massive amounts of side information is not guaranteed to improve performance relative to simpler multitask learning. Our results emphasize that multitask effects are highly dataset-dependent, suggesting the use of dataset-specific models to maximize overall performance.