aiqbench analysis
This arXiv study tests whether unlabeled molecules from large chemical databases like ChEMBL can improve semi-supervised QSAR models. It compares self-training, co-training, and label propagation across four sampling strategies on 10 datasets. The main finding is that selecting a smaller set of chemically distant molecules can matter more than raw unlabeled set size, which is useful for ML practitioners working on label-scarce scientific domains.
Not All Unlabeled Data Are Equal: Unlabeled Data Selection for Semi-Supervised QSAR Modeling with Large Chemical Repositories
arXiv:2610.22025v1 Announce Type: new 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.
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.
Machine Learning (cs.LG)
arXiv:2610.22025 [cs.LG]
(or arXiv:2610.22025v1 [cs.LG] for this version)
https://doi.org/10.48550/arXiv.2610.22025 Focus to learn more arXiv-issued DOI via DataCite (pending registration)
arXivLabs is a framework that allows collaborators to develop and share new arXiv features directly on our website.
Both individuals and organizations that work with arXivLabs have embraced and accepted our values of openness, community, excellence, and user data privacy. arXiv is committed to these values and only works with partners that adhere to them.
Have an idea for a project that will add value for arXiv's community? Learn more about arXivLabs.
Full text under the publisher's open licence · 319 words
Related coverage
Label-Efficient Deep Learning for ECG Delineation: A Multi-Dataset Benchmark against Widely Used Delineation Tools
arXiv · Oct 7, 2026
The art and science of hyperparameter optimization on Amazon Nova Forge
AWS · Jun 2, 2026
J++ Lens: Jacobian Filtering Enables More Faithful Workspace Lenses
LessWrong · Oct 8, 2026
Semi-Supervised Learning under Spatially Biased Sampling
arXiv · Sep 9, 2026