Task-specific pre-training for molecular property prediction
Briefings in Bioinformatics · 2026
DOI: 10.1093/bib/bbag010 ↗Kurz gefasst
Fachartikel in der Zeitschrift Briefings in Bioinformatics, 2026. Verfasst von Wenbo Zhang und 5 weiteren Autor:innen. Beteiligte Einrichtung: School of Computer Science and Technology, Xidian University, 266 Xinglong Section of Xifeng Road, Xi'an 710126, Shaanxi, China. Hauptthemen laut MEDLINE: Drug Discovery, Deep Learning, Computational Chemistry. Die Arbeit wurde bisher 2-mal zitiert (Crossref) und nennt 37 Literaturangaben. Der Volltext ist frei zugänglich (Open Access, CC BY-NC 4.0). Gefördert von: National Key R&D Program of China, Natural Science Foundation of Sichuan Province, Science and Technology Major Project of Sichuan Province.
Abstract (Originalsprache)
Molecular property prediction is a critical task in computational chemistry and drug discovery. While deep learning has advanced this field, the increasing complexity of models contrasts with the scarcity of labeled data, leading to severe overfitting and limited generalization. In this paper, we propose TasProp, a task-specific pre-training strategy for molecular property prediction, particularly for the scenarios with small labeled datasets. To learn a robust molecular representation, TasProp first projects both labeled and unlabeled data into a unified latent space. Then, we introduce a task-specific contrastive loss that aligns closely with the final prediction task and apply it to the labeled data. This contrastive loss encourages the model to learn more cohesive and distinguishable molecular representations corresponding to property categories, which in turn, enhances the model’s performance on downstream property prediction tasks. Additionally, we propose a novel data augmentation method, accompanied by a theoretical analysis, to mitigate the challenge of labeled data scarcity. With the task-specific pre-training and augmented data, TasProp outperforms the state-of-the-art methods on many molecular property prediction tasks, including three publicly available datasets and two curated datasets related to anesthesiology. Furthermore, we provide an interactive web resource to facilitate model exploration and application, allowing users to easily predict the properties of input molecules online.
Schlagwörter
MeSH-Schlagwörter (MEDLINE)
Stichwörter der Autor:innen
Autor:innen und Institutionen
- School of Computer Science and Technology, Xidian University, 266 Xinglong Section of Xifeng Road, Xi'an 710126, Shaanxi, China.
- College of Computer Science, Sichuan University, No. 24 South Section 1, Yihuan Road, Chengdu 610065, Sichuan, China.
- Laboratory of Anesthesia and Critical Care Medicine, Department of Anesthesiology, Translational Neuroscience Center, West China Hospital, Sichuan University, No. 17, Section 3, Renmin South Road, Chengdu 610041, Sichua…
ORCID-Profile: Wenbo Zhang, Xianggen Liu
Förderung
- National Key R&D Program of China
- Natural Science Foundation of Sichuan Province
- Science and Technology Major Project of Sichuan Province
- Fundamental Research Funds for the Central Universities
- Sichuan Province Engineering Technology Research Center of Broadband Electronics Intelligent Manufacturing
Zitieren
APA
Zhang, W., Wang, Y., Liu, J., Ke, B., Lv, J., & Liu, X. (2026). Task-specific pre-training for molecular property prediction. Briefings in Bioinformatics. https://doi.org/10.1093/bib/bbag010
Vancouver
Zhang W, Wang Y, Liu J, Ke B, Lv J, Liu X. Task-specific pre-training for molecular property prediction. Briefings in Bioinformatics. 2026. doi:10.1093/bib/bbag010
BibTeX
@article{Zhang2026,
author = {Zhang, Wenbo and Wang, Yihui and Liu, Jin and Ke, Bowen and Lv, Jiancheng and Liu, Xianggen},
title = {Task-specific pre-training for molecular property prediction},
journal = {Briefings in Bioinformatics},
year = {2026},
doi = {10.1093/bib/bbag010}
}
EUVIMED