It can be hard to get several international Big Pharma companies to work in harmony.
With an astounding amount of preclinical data and research being generated among them, it’s easy to see how not all of it may be used optimally, whether due to resource constraints or the urge to keep it closely held for competitive reasons. But were they to join a collaborative chorus, scientists could potentially make new discoveries using old data.
Enter MELLODDY. Short for Machine Learning Ledger Orchestration for Drug Discovery, the three-year project from the Innovative Medicines Initiative hopes to build an AI platform capable of ingesting large libraries of proprietary information while protecting its confidentiality.
Using a blockchain-based infrastructure, owners would retain control of their chemical and biological data across the duration of the project, which aims to accelerate drug research in any disease area by identifying promising molecules for development. The R&D work itself would be spread out among the participants.
The public-private consortium includes at least 10 Big Pharma companies and their European divisions: Amgen Research, based in Munich; Astellas Pharma’s location in the Netherlands; AstraZeneca’s Swedish subsidiary; and Janssen Pharmaceutica in Belgium, as well as Bayer, Boehringer Ingelheim, GlaxoSmithKline, Institut de Recherches Servier, Merck KGaA and Novartis.
Together, they’ve signed up to provide over a billion data points relevant to drug development from their chemical libraries, according to the IMI, as well as hundreds of terabytes of image data annotating the biological effects of over 10 million different small molecules.
To help them process it, MELLODDY includes a handful of data science and AI companies—including Iktos, Loodse, Owkin France, Substra and Nvidia Switzerland, as well as university partners and public research groups.
“The goal is to harness the collective knowledge of the consortium in a platform containing amongst others multi-task predictive machine learning algorithms incorporating an extended privacy management system, to identify the most effective compounds for drug development, while protecting the intellectual property rights of the consortium contributors,” Owkin project coordinator Mathieu Galtier said in a statement after the project launched this past June.
The data wouldn’t be pooled together in the traditional sense—in fact, none of it would leave the companies’ respective servers. Instead, nonsensitive models would be shared among the members, harvested by machine learning algorithms, and coordinated by the project’s central dispatcher.
“The exact chemistry or assays are not disclosed, but the partners can build models that are informed by the chemical and biological spaces of the combination of the other partners,” Hugo Ceulemans, MELLODDY project leader and scientific director of discovery data sciences at Janssen, told the Lancet.