Limited training data and a lack of diversity limit the performance and applicability domain of ADMET prediction models.
Predicting ADMET endpoints is challenging due to:
Process complexity
Low-throughput experimental assays
Limited chemically relevant data, especially in early-stage research
Only few pharmaceutical and biotech companies have large and diverse enough proprietary datasets to train robust QSAR models of sufficient quality.
The solution
Biopharmaceutical network for collaborative training of ADMET models while protecting data confidentiality.
Gain access to orders of magnitude more and more diverse data through collaboration.
Apheris enables pharma & biotech companies as well as academic researchers to train ADMET models on their proprietary data without the need for data sharing.
Public data lacks both quantity and chemical diversity, limiting ADMET models' predictive power and applicability domain. By joining a federation-powered ADMET Consortium, pharma companies and biotech can address this challenge and gain access to better models. Apheris has the expertise and product to enable secure, federated collaborations that protect everyone's IP and data privacy.
Learn more about the setup, the datasets contributed, and the models used in the ADMET Network
Powering the AISB Consortium to Revolutionize AI Drug Discovery
Apheris provides the tech layer for the Artificial Intelligence Structural Biology (AISB) Consortium, an unprecedented collaboration among AbbVie, Boehringer Ingelheim, Johnson & Johnson, Sanofi and Takeda aimed at transforming AI drug discovery. State-of-the-art AI models will be trained and evaluated on unique data from multiple biopharma companies without exposing proprietary information.