---
title: "GRASP: Graph Transformer Pretrained on ZINC20 and Supervised on ChEMBL for Molecular Property Pred"
id: "biorxiv-18-grasp-graph-representation-learning-with-assay-supervision-for-molecular"
canonical_url: "https://medichelpline.com/clinical-feed/biorxiv-18-grasp-graph-representation-learning-with-assay-supervision-for-molecular"
content_type: "clinical_feed_article"
specialty: "Pharmacology"
source_name: "bioRxiv (Biomedical Preprints)"
source_url: "https://www.biorxiv.org/content/10.64898/2026.09.28.755019v1?rss=1"
published_at: "2026-10-07T12:00:00.000Z"
evidence_level: "Verified Feed"
license: "CC-BY-NC-4.0 / Informational Use"
---
# GRASP: Graph Transformer Pretrained on ZINC20 and Supervised on ChEMBL for Molecular Property Pred
## Provenance & Clinical Metadata
- **Canonical URL:** https://medichelpline.com/clinical-feed/biorxiv-18-grasp-graph-representation-learning-with-assay-supervision-for-molecular
- **Specialty:** [Pharmacology](https://medichelpline.com/clinical-feed/pharmacology.md)
- **Primary Source:** bioRxiv (Biomedical Preprints)
- **Source URL:** [Original Journal Publication](https://www.biorxiv.org/content/10.64898/2026.09.28.755019v1?rss=1)
- **Published At:** 2026-10-07T12:00:00.000Z
- **Evidence Rating:** Verified Feed
## Executive GIST (TL;DR)
- The preprint introduces **GRASP**, a 93.5M-parameter graph Transformer designed for molecular representation learning that combines large-scale structural pretraining with intermediate assay supervision. - GRASP is pretrained by learning molecular structure using a **replaced-token detection** objective on 1.54 billion molecules from the **ZINC20** collection. - After structural pretraining, GRASP is adapted to experimental bioactivity by supervising on sparse measurements from 642 **ChEMBL** assays, i.e., an intermediate bioactivity adaptation stage. - On 23 cluster-held-out OpenADMET endpoints, GRASP achieved a mean **MAE** of 0.374, outperforming CheMeleon (0.383), ECFP4-LightGBM (0.408), and Chemprop (0.440). - The ChEMBL adaptation step reduced the mean MAE from 0.407 (post-structure pretraining only) to 0.374, improving 21 of 23 OpenADMET endpoints and demonstrating the benefit of intermediate assay supervision. - On 22 TDC ADMET tasks, GRASP provided a better point estimate than CheMeleon on 13 tasks, indicating competitive performance across multiple ADMET benchmarks. - The authors frame these results as support for a staged molecular representation learning approach: abundant molecular structure and sparse experimental measurements provide complementary supervision across training stages. - This work is reported as a preprint that has not undergone peer review; competing interests were declared as none by the authors.
## Clinical Analysis & Structured Key Points
GRASP: Graph Representation Learning with Assay Supervision for Molecular Properties | bioRxiv Skip to main content New Results GRASP: Graph Representation Learning with Assay Supervision for Molecular Properties View ORCID Profile Satya Pratik Srivastava , View ORCID Profile Rohan Gorantla , Sharath Krishna Chundru , Harshit Singh , View ORCID Profile Antonia S. J. S. Mey , View ORCID Profile Rajeev Kumar Singh doi: https://doi.org/10.64898/2026.09.28.755019 Satya Pratik Srivastava 1 Shiv Nadar University , Delhi-NCR, India Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Satya Pratik Srivastava Rohan Gorantla 2 School of Informatics, University of Edinburgh , Edinburgh EH8 9AB, U.K. 3 EaStCHEM School of Chemistry, University of Edinburgh , Edinburgh EH9 3FJ, U.K. Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Rohan Gorantla Sharath Krishna Chundru 1 Shiv Nadar University , Delhi-NCR, India Find this author on Google Scholar Find this author on PubMed Search for this author on this site Harshit Singh 1 Shiv Nadar University , Delhi-NCR, India Find this author on Google Scholar Find this author on PubMed Search for this author on this site Antonia S. J. S. Mey 3 EaStCHEM School of Chemistry, University of Edinburgh , Edinburgh EH9 3FJ, U.K. Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Antonia S. J. S. Mey For correspondence: rajeev.kumar{at}snu.edu.in Antonia.Mey{at}ed.ac.uk Rajeev Kumar Singh 1 Shiv Nadar University , Delhi-NCR, India Find this author on Google Scholar Find this author on PubMed Search for this author on this site ORCID record for Rajeev Kumar Singh For correspondence: rajeev.kumar{at}snu.edu.in Antonia.Mey{at}ed.ac.uk Abstract Full Text Info/History Metrics Preview PDF Abstract Molecular structures are abundant, while experimental bioactivity is sparse and distributed across assays. We introduce GRASP, a 93.5M-parameter graph Transformer that learns from these sources in sequence. GRASP first learns molecular structure through replaced-token detection on 1.54 billion ZINC20 molecule presentations, then adapts to sparse bioactivity across 642 ChEMBL assays. Across 23 cluster-held-out OpenADMET endpoints, GRASP achieves a mean MAE of 0.374, compared with 0.383 for CheMeleon, 0.408 for ECFP4-LightGBM, and 0.440 for Chemprop. ChEMBL adaptation reduces mean MAE from 0.407 to 0.374 and improves 21 of 23 endpoints, showing the value of intermediate bioactivity supervision. On 22 TDC ADMET tasks, GRASP gives a better point estimate than CheMeleon on 13 tasks. These results support staged molecular representation learning, where abundant molecular structure and sparse experimental measurements provide complementary supervision at different stages of training. Competing Interest Statement The authors have declared no competing interest. Copyright The copyright holder for this preprint is the author/funder, who has granted bioRxiv a license to display the preprint in perpetuity. It is made available under a CC-BY-NC 4.0 International license . Back to top Previous Next Posted October 02, 2026. Download PDF Email Thank you for your interest in spreading the word about bioRxiv. NOTE: Your email address is requested solely to identify you as the sender of this article. Your Email * Your Name * Send To * Enter multiple addresses on separate lines or separate them with commas. You are going to email the following GRASP: Graph Representation Learning with Assay Supervision for Molecular Properties Message Subject (Your Name) has forwarded a page to you from bioRxiv Message Body (Your Name) thought you would like to see this page from the bioRxiv website. Your Personal Message CAPTCHA This question is for testing whether or not you are a human visitor and to prevent automated spam submissions. Share GRASP: Graph Representation Learning with Assay Supervision for Molecular Properties Satya Pratik Srivastava , Rohan Gorantla , Sharath Krishna Chundru , Harshit Singh , Antonia S. J. S. Mey , Rajeev Kumar Singh bioRxiv 2026.09.28.755019; doi: https://doi.org/10.64898/2026.09.28.755019 Share This Article: Copy Citation Tools GRASP: Graph Representation Learning with Assay Supervision for Molecular Properties Satya Pratik Srivastava , Rohan Gorantla , Sharath Krishna Chundru , Harshit Singh , Antonia S. J. S. Mey , Rajeev Kumar Singh bioRxiv 2026.09.28.755019; doi: https://doi.org/10.64898/2026.09.28.755019 Citation Manager Formats BibTeX Bookends EasyBib EndNote (tagged) EndNote 8 (xml) Medlars Mendeley Papers RefWorks Tagged Ref Manager RIS Zotero Tweet Widget Facebook Like Google Plus One Subject Area Bioinformatics Subject Areas All Articles Animal Behavior and Cognition (8093) Biochemistry (18912) Bioengineering (15071) Bioinformatics (44867) Biophysics (22789) Cancer Biology (19922) Cell Biology (27132) Clinical Trials (138) Developmental Biology (14064) Ecology (21230) Epidemiology (2067) Evolutionary Biology (25633) Genetics (16265) Genomics (23664) Immunology (18846) Microbiology (42855) Molecular Biology (18218) Neuroscience (94301) Paleontology (708) Pathology (3011) Pharmacology and Toxicology (5137) Physiology (8197) Plant Biology (16138) Scientific Communication and Education (2105) Synthetic Biology (4604) Systems Biology (10317) Zoology (2405)
## Related Clinical Research

- [Community pharmacy telephone counseling often leads to unnecessary oral emergency contraception re](https://medichelpline.com/clinical-feed/plos-one-20-do-community-pharmacies-unnecessarily-recommend-oral-emergency-contraception-a.md)
- [Biocatalytic production of Danshensu from L‑Tyrosine using engineered Escherichia coli](https://medichelpline.com/clinical-feed/plos-one-1-multienzyme-cascade-biotransformation-of-l-tyrosine-to-danshensu-by-engineered.md)
- [Callio Therapeutics hires Chris Hassig as chief scientific officer](https://medichelpline.com/clinical-feed/stat-news-6-stat-up-and-down-the-ladder-the-latest-comings-and-goings.md)
- [U.S. Treasury Proposes Restricting Investment in Chinese Biotech Under COINS Rule](https://medichelpline.com/clinical-feed/stat-news-6-stat-treasury-proposes-major-restrictions-to-u-s-investment-in-chinese-biotech.md)
- [UnibindMD: AI-plus-MD Platform for De Novo Peptide Binder Design](https://medichelpline.com/clinical-feed/biorxiv-16-unibindmd-an-integrated-ai-and-molecular-dynamics-platform-for-de-novo-peptide.md)

## Navigation
- [← Back to Pharmacology Feed](https://medichelpline.com/clinical-feed/pharmacology.md)
- [← All Clinical Specialties](https://medichelpline.com/clinical-feed.md)
## Medical & Regulatory Disclaimer

> [!CAUTION]
> MedicHelpline content is structured for research, educational, and professional discovery purposes. It does not constitute individual medical advice, clinical diagnosis, or treatment recommendations.
> Always verify dosing, contraindications, and regulatory alerts against official product labeling and primary regulatory sources before clinical decision-making.