A Computational Community Blind Challenge on Pan-Coronavirus Drug Discovery Data

Area of Science:

• Drug Discovery
• Computational Chemistry
• Machine Learning

Background:

• Computational blind challenges are crucial for unbiased scientific progress.
• The AI-driven Structure-enabled Antiviral Platform Discovery Consortium (SAPCDC) focuses on pan-coronavirus antiviral discovery.
• OpenADMET project and Polaris platform facilitate collaborative drug discovery initiatives.

Purpose of the Study:

• To report outcomes and insights from an open science community blind challenge in computational drug discovery.
• To assess computational methods for predicting biochemical potency and crystallographic ligand poses against SARS-CoV-2 and MERS-CoV Mpro.
• To evaluate multiple ADMET (Absorption, Distribution, Metabolism, Excretion, Toxicity) assay endpoints.

Main Methods:

• Global participants from academia and industry developed and applied computational methods.
• Previously undisclosed experimental drug discovery data sets served as benchmarks.
• Submissions were evaluated across multiple tasks and compounds, establishing performance leaderboards.
• Meta-analyses were conducted to assess methodological strengths and weaknesses.

Main Results:

• Performance leaderboards were established for various computational tasks.
• Key insights into methodological strengths and common pitfalls were identified.
• The study provides a foundation for best practices in real-world machine learning evaluation.

Conclusions:

• The challenge advanced reproducible, trustworthy, and high-impact computational methods in drug discovery.
• Best practices and pitfalls in future blind challenge design and execution were explored.
• Next-generation platforms like Polaris enable rigorous challenge design and community engagement.