Dr Jean-Philippe RAMEAU
Head of Molecular Design Science France
Sanofi Research & Development
A.I. Driven Drug Discovery
Drug discovery is a complex problem, finding a synthetic chemical (Synthetic Modalities) that fulfill all the requirement to become a drug is a challenge. Over the years the process of discovery has evolved from observation and serendipity to more data-driven and then rationally data driven process. Even if we claim from decade that the rational drug design is a core activity of drug discovery the evolution of computing capabilities and the development of new algorithm especially in the field of Artificial Intelligence have given a boost to this domain. Today the Artificial intelligence is under the spotlight in drug discovery and some recent developments in machine learning especially neural networks brought interesting opportunities for the chemical space exploration.
In this talk I will present some examples of Machine-learning capabilities, from the single point exploration of chemical space (predictive models)) to the optimized exploration of the chemical space provided by the denovo design by reinforcement learning. Machine learning can also be used on top of other more traditional physics-based technologies to increase the capability and access to numbers not usually accessible i.e., Virtual-Screening or combined with Free-Energy Perturbation. Combination of machine learning and first principal technologies gives access to a powerful mixed technology that push the boundaries of research. A lot is expected from of these technologies to explore the “druggability” of some difficult biological systems.
Today these algorithms are used daily to support our research portfolio but still the most important operation remain the decision-making process. Although assisted by machine especially using multi-parametric optimization, the human remains at the center. In this respect Artificial Intelligence provides a real improvement of our “In-Silico toolbox” but on a project perspective the intelligence lies into the team. Today we prefer to refer to “Augmented Medchem” or “Augmented Intelligence” when speaking of this shift of ways of working. But still drug discovery remains a complex problem…