Largest-ever pharma industry artificial intelligence collaboration utilizing a federated model holds potential benefits for drug discovery without compromising IP, landmark project finds
The MELLODDY project has leveraged the world’s largest collection of small molecules with known biochemical or cellular activities to enable more accurate predictive modeling. Using federated and privacy-preserving machine learning, the results may enable increasing efficiencies in drug discovery.
July 13, 2022 – A landmark project that enabled 10 pharmaceutical companies and seven technology and academic strategic partners to work together has successfully demonstrated that collaborating in artificial intelligence (AI) for drug discovery is possible at industrial scale. The results indicate that collaborative modeling among pharmaceutical partners holds potential benefits for the discovery of new drugs.
MELLODDY, a three-year project co-funded (€18 million) by the European Union and partnering EFPIA companies under the Innovative Medicines Initiative (IMI2), has achieved its ambition of building a secure platform for privacy-preserving and federated learning. This novel AI framework avoids the need for competitive data and models to ever leave the owner’s custody while still allowing collaborative machine learning that can train and evaluate drug discovery predictive models.
The results show that its operation at scale yields improvements across all pharmaceutical partners in the predictive performance of collaboratively trained models over single partner models. Models that more accurately predict the pharmacological and toxicological activities of molecules better support the decision-making process of which candidate drug molecules to make and test.
Hugo Ceulemans, Project Lead, Senior Scientific Director Drug Discovery Data Sciences at Janssen Pharmaceutica NV, said:
“Throughout the pharmaceutical industry, increasingly powerful machine learning approaches are leveraging ever more data and insights to better focus and accelerate the real-world experiments and studies that bring life-saving drugs to patients. MELLODDY has positioned federated learning in this context, demonstrating the concrete feasibility and benefits of collaborative yet competitive modeling.”
Mathieu Galtier, Project Coordinator, Chief Data & Platform Officer at Owkin, said:
“This is a massive win for drug discovery and ultimately patients. For years, it was assumed that competing pharmaceutical companies could never work together at scale to discover the next generation of drugs. But our three-year project showed that it is not only possible – it is also more effective than going at it alone. Crucially, we demonstrated that federated learning protects commercially sensitive data, meaning that no one loses out when collaborating.”
Launched in 2019, MELLODDY announced the successful development and operation of a secure platform for federated learning without sharing each partner’s proprietary data and models or compromising their security and confidentiality in 2020. In 2021, MELLODDY shared the first-ever demonstration of the predictive performance benefits of federated learning in drug discovery. Today, the consortium announces its scientific results following the completion of its ambitious three-year endeavour.
Operating at an unprecedented scale, the MELLODDY platform trained models on billions of industrial experimental datapoints documenting the behavior of more than 20 million chemical small molecules in over 40,000 biological assays.
Results
To evaluate model improvement while accounting for the wide variety in assays, for each multi-partner model, the proximity to a perfect score was related to that of the corresponding single partner model.
Across the pharmaceutical partners, collaborative models were typically 4% better at categorizing molecules as either pharmacologically or toxicologically active, or not active. The typical multi-partner model also showed a 10% increase in its applicability domain, its ability to yield confident predictions when applied to new types of molecules. Finally, the collaborative models were typically 2% better at estimating values of toxicological and pharmacological activities. Performance gains proved more prominent for the subset of assays relating to pharmacokinetics and toxicology, and for assays with ongoing data acquisition. Collectively, these results show improvements to predictive models that support the drug discovery process.
As MELLODDY comes to its official close, the technology developed as part of this project will be extended to new fields in healthcare to facilitate privacy-preserving collaboration in AI. While the consortium focused on the domain of small molecule drug discovery, its approach has the potential to benefit other areas in the development pipeline, including areas of interest such as biologics, histology, and genomics. Parties interested in initiating or joining new consortia for machine learning-driven research in healthcare are invited to contact Wilfried Dang, Director of Business Development, Discovery and Early R&D at Owkin (wilfried.dang@owkin.com).
About MELLODDY
The consortium was funded by the Innovative Medicines Initiative (IMI), a partnership between the European Union and the European pharmaceutical industry, represented by the European Federation of Pharmaceutical Industries and Associations (EFPIA). Consortium members included 10 pharmaceutical companies (Amgen; Astellas; AstraZeneca; Bayer; Boehringer Ingelheim; GSK; Institut de Recherches Servier; Janssen Pharmaceutica NV; Merck KGaA; and Novartis), five technology partners (Iktos; Kubermatic; NVIDIA; Owkin; and Substra Foundation), and two academic partners (Budapest University of Technology and Economics; KU Leuven). Amazon Web Services has advised on optimal leveraging of their technology for computational execution and generously contributed to the project’s computation budget. The project involved more than 100 experts in computational chemistry, data science, algorithmics, software engineering and deployment, IT operations and security, and project management.
Contact
MELLODDY Communications: Alexander Blackburn – Owkin (alexander.blackburn@owkin.com)
Acknowledgement
This project has received funding from the Innovative Medicines Initiative 2 Joint Undertaking under grant agreement No 831472. This Joint Undertaking receives support from the European Union’s Horizon 2020 research and innovation programme and EFPIA Companies.
Disclaimer
This communication reflects the views of the authors and neither IMI nor the European Union, EFPIA or any Associated Partners are liable for any use that may be made of the information contained herein