Federated AI Advances in Binding Affinity and ADMET Prediction for Drug Discovery

Apheris launched the ADMET Network, a federated learning platform uniting pharma companies like Lundbeck, Orion Pharma, Recursion, and Servier to train privacy-preserving AI models for ADMET predictions using 80% of their proprietary data.1

SandboxAQ released AQAffinity on January 20, 2026, a structure-free binding affinity prediction model built on OpenFold3, enabling rapid evaluation of drug candidates without experimental structures.2

The OpenFold Consortium, involving Apheris collaborations with AbbVie, J&J, and others, fine-tunes OpenFold3 for protein-ligand interactions in a confidential environment.12

2026 predictions highlight maturing protein structure and binding affinity prediction models, though they require hybrid AI-physics approaches for drug discovery success.3

Sources:

1. https://www.genengnews.com/topics/artificial-intelligence/admet-predictions-get-ai-boost-federated-data-network-unites-pharma/

2. https://www.prnewswire.com/news-releases/sandboxaq-supercharges-openfold3-with-aqaffinity-high-speed-structure-free-drug-potency-prediction-302665496.html

3. https://www.drugtargetreview.com/article/192962/ai-in-drug-discovery-predictions-for-2026/