Mapping the avoid-ome: a systematic open-science approach to predictive ADMET
5 affiliations
- 1Department of Bioengineering and Therapeutic Sciences, University of California San Francisco, San Francisco, California, USA and Quantitative Biosciences Institute, University of California San Francisco, San Francisco, California, USA. [email protected]
- 2Octant, Inc, Emeryville, CA, USA
- 3Computational and Systems Biology Program, Sloan Kettering Institute, Memorial Sloan Kettering Cancer Center, New York, NY, USA
- 4Disruptive Biomedical LLC, Holliston, MA, USA
- 5OpenADMET, Boston, MA, USA. [email protected]
Abstract
Drug discovery faces significant obstacles posed by unpredictable pharmacokinetic and safety properties, necessitating a complex, multiparameter optimization process. ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) issues remain the primary reason for failure in drug development, with over 90% of discovery compounds failing basic ADME standards and unexpected ADMET problems causing around 30% of clinical setbacks. Conventional approaches that mainly rely on broad molecular properties offer limited guidance because they lack detailed insight into the atomic-level interactions between drugs and the body’s complex systems. Transforming small-molecule drug discovery requires a systematic, detailed understanding of the "Avoid-ome": the broad set of proteins that influence ADME and toxicity characteristics. The Avoid-ome includes a finite, manageable set of enzymes, transporters, receptors, and channels, which must be treated as "anti-targets" and avoided during the design process. OpenADMET (https://openadmet.org), an international open-science initiative, aims to fill the critical ADMET data gap by creating pre-competitive, open datasets covering metabolism, transport, distribution, and toxicity. The initiative is developing platforms that make compound synthesis, measurements (using technologies like scaled mass spectrometry and synthetic biology), and data analysis more affordable and capable of high-throughput processing. This approach uses high-throughput structural biology to ensure models are based on mechanistic, atomistic understanding, helping to clarify the structural basis of outliers, species differences, and genetic variation. Additionally, an active learning workflow is used across diverse chemical spaces to select compounds that are the most informative for building generalizable predictive models. OpenADMET includes blind community challenges, inspired by CASP and SAMPL, to evaluate predictive models with unreleased data, encouraging rigorous assessment and ongoing improvement within the research community. By systematically studying the Avoid-ome and creating open, structural, and mechanistic datasets, OpenADMET establishes a foundation for a new era of rational drug design, demonstrating that the most effective way to improve drug discovery is to stop avoiding the Avoid-ome and instead study it directly.
Related
- Systematic evaluation of LLM-based zero-shot model construction in chemistry and materials science: predictive performance and response reliability
- Systematic Evaluation of Mixture-Design for Multi-Oxide Magnesium Aluminosilicate Glass Development: Model Comparison, Modifier Interactions, and Predictive Limitations
- Systematic mapping of bimetallic catalyst combinations for floating catalyst chemical vapor deposition growth of single-walled carbon nanotubes
- Integrating dft-based spectroscopic analysis, molecular docking, admet prediction, and natural biosorbent adsorption studies of flupyradifurone
- Accelerating Lead Optimization away from Anti-Targets: Navigating the Structural Complexity of Promiscuous ADMET Proteins
BibTeX
@article{fraser2026mapping,
title = {{Mapping the avoid-ome: a systematic open-science approach to predictive ADMET}},
author = {James S. Fraser and Steven Edgar and L. Naomi Handly and Sriram Kosuri and John D. Chodera and Mark A. Murcko and W. Patrick Walters},
journal = {Nature Communications},
year = {2026},
month = may,
doi = {10.1038/s41467-026-73410-8},
url = {https://doi.org/10.1038/s41467-026-73410-8},
note = {CC BY-NC-ND},
}
RIS
TY - JOUR
TI - Mapping the avoid-ome: a systematic open-science approach to predictive ADMET
AU - James S. Fraser
AU - Steven Edgar
AU - L. Naomi Handly
AU - Sriram Kosuri
AU - John D. Chodera
AU - Mark A. Murcko
AU - W. Patrick Walters
PY - 2026
DA - 2026/05/25
JO - Nature Communications
DO - 10.1038/s41467-026-73410-8
UR - https://doi.org/10.1038/s41467-026-73410-8
AB - Drug discovery faces significant obstacles posed by unpredictable pharmacokinetic and safety properties, necessitating a complex, multiparameter optimization process. ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) issues remain the primary reason for failure in drug development, with over 90% of discovery compounds failing basic ADME standards and unexpected ADMET problems causing around 30% of clinical setbacks. Conventional approaches that mainly rely on broad molecular properties offer limited guidance because they lack detailed insight into the atomic-level interactions between drugs and the body’s complex systems. Transforming small-molecule drug discovery requires a systematic, detailed understanding of the "Avoid-ome": the broad set of proteins that influence ADME and toxicity characteristics. The Avoid-ome includes a finite, manageable set of enzymes, transporters, receptors, and channels, which must be treated as "anti-targets" and avoided during the design process. OpenADMET (https://openadmet.org), an international open-science initiative, aims to fill the critical ADMET data gap by creating pre-competitive, open datasets covering metabolism, transport, distribution, and toxicity. The initiative is developing platforms that make compound synthesis, measurements (using technologies like scaled mass spectrometry and synthetic biology), and data analysis more affordable and capable of high-throughput processing. This approach uses high-throughput structural biology to ensure models are based on mechanistic, atomistic understanding, helping to clarify the structural basis of outliers, species differences, and genetic variation. Additionally, an active learning workflow is used across diverse chemical spaces to select compounds that are the most informative for building generalizable predictive models. OpenADMET includes blind community challenges, inspired by CASP and SAMPL, to evaluate predictive models with unreleased data, encouraging rigorous assessment and ongoing improvement within the research community. By systematically studying the Avoid-ome and creating open, structural, and mechanistic datasets, OpenADMET establishes a foundation for a new era of rational drug design, demonstrating that the most effective way to improve drug discovery is to stop avoiding the Avoid-ome and instead study it directly.
SN - 2041-1723
ER -