Unlabeled chemical data quality matters more than quantity in activity prediction
Not All Unlabeled Data Are Equal: Unlabeled Data Selection for Semi-Supervised QSAR Modeling with Large Chemical Repositories
Machine Learning
Summary
Predicting how molecules behave biologically often needs expensive experiments, so scientists have only small sets of known examples. The authors studied how to use many unlabeled molecules from big databases to improve these predictions without extra experiments. They found that choosing a small, diverse set of unlabeled molecules helps predictions more than just using lots of data. This challenges the usual idea that more unlabeled data always makes machine learning better.
What this means in practice
- •For drug discovery teams: Improve molecular activity prediction models by selecting small, chemically diverse unlabeled datasets from large chemical databases for semi-supervised learning.
- •For chemical data engineers: Design pipelines that prioritize chemically distant unlabeled molecules to enhance QSAR semi-supervised learning effectiveness without increasing dataset size.
Authors
Oliver Majer, Saso Dzeroski, Jurica Levatic
Abstract
Semi-supervised learning (SSL) is particularly attractive in domains where labels are expensive to obtain, while unlabeled examples are readily available. Quantitative Structure-Activity Relationship (QSAR) modeling, which predicts molecular biological activity from chemical structure, is one such domain. Obtaining activity measurements requires costly wet-lab experiments, resulting in small labeled datasets, while chemical repositories such as ChEMBL contain millions of unlabeled molecules. Despite this apparent suitability, the potential of large chemical databases to improve QSAR models via SSL remains largely unexplored. Furthermore, it is unclear how unlabeled data should be selected in realistic SSL settings with genuine unlabeled data. We conduct a systematic study of unlabeled-data selection using 10 QSAR datasets extracted from ChEMBL and a pool of more than 400,000 unlabeled molecules. To this end, we evaluate three established SSL algorithms (self-training, co-training, and label propagation) and formulate four unlabeled-data selection strategies based on random, nearest-neighbor, furthest-neighbor, and mixed sampling. Our results show that unlabeled data from large chemical repositories can improve QSAR models, with at least one SSL configuration outperforming the supervised baseline on each of the 10 datasets. Moreover, the most reliable improvements are generally achieved by selecting a relatively small number of chemically distant molecules. This suggests that the composition of the unlabeled set may be more important than its size, challenging the common premise that SSL necessitates an abundance of unlabeled data.