Assessing the state of the art in ADMET prediction: Lessons from the OpenADMET-ExpansionRx community blind challenge

@misc{castellanos2026assessing,
  title = {{Assessing the state of the art in ADMET prediction: Lessons from the OpenADMET-ExpansionRx community blind challenge}},
  author = {Maria A. Castellanos and Hugo MacDermott-Opeskin and Jon Ainsley and Jonathan A. Swain and Sean Colby and Devany West and Georgia Channing and Andrew Good and Elyse Bourque and Lakshminarayana Vogeti and Renato Skerlj and Tiansheng Wang and Mark Ledeboer and David Alencar Araripe and Rafal Bachorz and Amitesh Badkul and Maximilian Beckers and Jérémy Besnard and Olivier Beyens and Maicol Bissaro and Davide Boldini and Laetitia Breuil and Jackson W Burns and Lianjin Cai and Ruel Cedeno and Joanna Ceklarz and Xin Chen and Alan C. Cheng and Vishnu Vardhan Chundu and Vladimir Chupakhin and Thomas Coudrat and Simon Crouzet and Daniel Crusius and Wim Dehaen and Valeriia Fil and Robert Fraczkiewicz and Jozef Fülöp and Enrico Gandini and David Giganti and Anthony Gitter and Alec Glisman and Pawel Gniewek and Ryan Greenhalgh and Malte Grieswelle and Oleg Gromov and Sudhir Gupta and Leonard F. Haasbroek and Calum Hand and Xibing He and Dimitrios Iliadis and Abdul Haleem Mohamed Iliiyas and Eachan Johnson and Kalen Josifovski and Shivananda Kandagalla and Lukas Kerti and Oliver Koch and Vaishnavi Kodamala and Stephanie Labouille and Shaolong Lin and Dmitriy Makarov},
  howpublished = {ChemRxiv},
  year = {2026},
  month = sep,
  doi = {10.26434/chemrxiv.15009342/v1},
  url = {https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009342/v1},
  note = {CC BY},
}

TY  - UNPB
TI  - Assessing the state of the art in ADMET prediction: Lessons from the OpenADMET-ExpansionRx community blind challenge
AU  - Maria A. Castellanos
AU  - Hugo MacDermott-Opeskin
AU  - Jon Ainsley
AU  - Jonathan A. Swain
AU  - Sean Colby
AU  - Devany West
AU  - Georgia Channing
AU  - Andrew Good
AU  - Elyse Bourque
AU  - Lakshminarayana Vogeti
AU  - Renato Skerlj
AU  - Tiansheng Wang
AU  - Mark Ledeboer
AU  - David Alencar Araripe
AU  - Rafal Bachorz
AU  - Amitesh Badkul
AU  - Maximilian Beckers
AU  - Jérémy Besnard
AU  - Olivier Beyens
AU  - Maicol Bissaro
AU  - Davide Boldini
AU  - Laetitia Breuil
AU  - Jackson W Burns
AU  - Lianjin Cai
AU  - Ruel Cedeno
AU  - Joanna Ceklarz
AU  - Xin Chen
AU  - Alan C. Cheng
AU  - Vishnu Vardhan Chundu
AU  - Vladimir Chupakhin
AU  - Thomas Coudrat
AU  - Simon Crouzet
AU  - Daniel Crusius
AU  - Wim Dehaen
AU  - Valeriia Fil
AU  - Robert Fraczkiewicz
AU  - Jozef Fülöp
AU  - Enrico Gandini
AU  - David Giganti
AU  - Anthony Gitter
AU  - Alec Glisman
AU  - Pawel Gniewek
AU  - Ryan Greenhalgh
AU  - Malte Grieswelle
AU  - Oleg Gromov
AU  - Sudhir Gupta
AU  - Leonard F. Haasbroek
AU  - Calum Hand
AU  - Xibing He
AU  - Dimitrios Iliadis
AU  - Abdul Haleem Mohamed Iliiyas
AU  - Eachan Johnson
AU  - Kalen Josifovski
AU  - Shivananda Kandagalla
AU  - Lukas Kerti
AU  - Oliver Koch
AU  - Vaishnavi Kodamala
AU  - Stephanie Labouille
AU  - Shaolong Lin
AU  - Dmitriy Makarov
PY  - 2026
DA  - 2026/09/23
PB  - ChemRxiv
DO  - 10.26434/chemrxiv.15009342/v1
UR  - https://chemrxiv.org/doi/full/10.26434/chemrxiv.15009342/v1
AB  - Computational blind challenges are essential for advancing drug discovery, especially in research areas with limited public data, such as absorption, distribution, metabolism, excretion, toxicity (ADMET) prediction. In this paper, we present the results of our latest community blind challenge, organized in collaboration with Expansion Therapeutics (ExpansionRx). Participants were tasked with predicting nine distinct ADMET endpoints routinely evaluated during lead optimization of compounds in drug discovery programs. Following unprecedented engagement from the global community, this challenge provided valuable insights into the state of the art in machine learning for ADMET modeling. Evaluating submissions against a realistic, challenging dataset from active ExpansionRx drug discovery programs revealed that message-passing neural networks and chemical foundation models dominated among the top performers. Furthermore, the results highlighted the power of proprietary industry datasets and a critical need to optimize predictive workflows for different endpoints independently via multitasking and task-affinity grouping. We also demonstrate how hosting our open-source blind challenge infrastructure on a flexible platform like Hugging Face was pivotal for driving community engagement and visibility. Ultimately, blind challenges supported by robust infrastructure, widespread dissemination, and rigorous post-challenge analysis are fundamental to maximizing the utility of public data releases. We anticipate that these community benchmarks will continue to drive the advancement of ADMET models, enabling their deployment in both early- and late-stage drug discovery.
ER  -