BEACON Launches to Unite AI Benchmarking Across Biology and Drug Discovery

A new international consortium, BEACON (Benchmarking, Evaluation, and Assessment Consortium for Science), has been launched to align and expand benchmarking efforts across AI-driven biology and medicine. The initiative brings together established critical assessment groups and open science organizations to develop shared evaluation standards, run community challenges, and create an open platform for validating predictive models in disease research and small-molecule discovery.

BEACON is developing a structured framework to assess how computational methods perform in complex and emerging biological systems, with the aim of improving research reproducibility in AI-driven science and promoting methodological clarity.

The consortium will also focus on benchmarking foundation models and AI agents used in biology, medicine, and drug discovery. Planned activities include convening a benchmarking think tank and operating an open platform for coordinated evaluations and shared datasets.

Conscience, a non-profit organization focused on open and collaborative drug discovery, will act as central coordinator.

Conscience runs other initiatives, such as the Developing Medicines through Open Science (DMOS) program, which provides staged funding from early lead identification through Phase 1–2 clinical development, with awards reaching up to $2M per project. The organisation also stewards the CACHE (Critical Assessment of Computation Hit-finding Experiments) Challenges, which benchmark computational hit discovery through prospective prediction and experimental validation with unrestricted public release of resulting data.

BEACON consortium founding members include:

• CASP (Critical Assessment of Structure Prediction), known for its community-wide protein structure prediction challenges.
• DREAM (Dialogue on Reverse Engineering Assessment and Methods), which organizes open competitions in systems biology and translational research.
• Sage Bionetworks, focused on open science and data-driven biomedical research.
• OpenADMET, which develops and evaluates predictive models for drug metabolism, pharmacokinetics, and toxicity.
• CACHE/Conscience (Critical Assessment of Computation Hit-finding Experiments), which benchmarks computational hit-finding methods for small-molecule discovery.

The consortium will be formally introduced to the broader community during a dedicated benchmarking session at the MAINFRAME Symposium on AI-Driven Small-Molecule Drug Discovery in Barcelona on March 18–19, 2026. That session is intended to outline BEACON’s framework for open evaluation and gather input from the machine learning and computational chemistry communities as the initiative defines its early priorities.

BEACON sits alongside a set of open, challenge-driven efforts that have shaped how AI and computational methods are evaluated in life sciences, including the Critical Assessment of Structure Prediction (CASP), a long-running community experiment that benchmarks 3D structure prediction methods on blind protein targets; the Critical Assessment of Genome Interpretation (CAGI), which runs periodic open challenges on predicting the impact of human genetic variants and related phenotypes; and OpenEBench, the ELIXIR benchmarking and technical monitoring platform that hosts community assessments for bioinformatics tools, web servers, and workflows, among other initiatives.

Topic: AI in Bio