Designing reward functions that guide generative molecular design (GMD) algorithms to desirable areas of chemical space is of critical importance in AI-driven drug discovery. Traditionally, this has been a manual and error-prone task; the selection of appropriate computational methods to approximate biological assays is challenging and the aggregation of computed values into a single score even more so, leading to potential reliance on trial-and-error approaches. We propose a novel approach for automated reward configuration that relies solely on experimental data, mitigating the challenges of manual reward adjustment on drug discovery projects. Our method achieves this by constructing a ranking over experimental data based on Pareto dominance over the multi-objective space, then training a neural network to approximate the reward function such that rankings determined by the predicted reward correlate with those determined by the Pareto dominance relation. We validate our method using two case studies. In the first study we simulate Design-Make-Test-Analyse (DMTA) cycles by alternating reward function updates and generative runs guided by that function. We show that the learned function adapts over time to yield compounds that score highly with respect to evaluation functions taken from the literature. In the second study we apply our algorithm to historical data from four real drug discovery projects. We show that our algorithm yields reward functions that outperform the predictive accuracy of human-defined functions, achieving an improvement of up to 0.4 in Spearman's correlation against a ground truth evaluation function that encodes the target drug profile for that project. Our method provides an efficient data-driven way to configure reward functions for GMD, and serves as a strong baseline for future research into transformative approaches for the automation of drug discovery.

AI is being used everywhere, including in drug discovery. There are many things AI drug discovery technology can do to further advance treatments for various diseases, but there are some things that AI cannot do. Understanding the applicability of AI predictions is essential. Without that understanding, it is easy to believe that AI alone can solve any problem autonomously when, in fact, researchers are also essential to the process. AI drug discovery is vital to the future of the pharmaceutical and healthcare industries.

Aqemia announced on October 19 that it had raised €30 million in a Series A financing round co-led by Eurazeo and Bpifrance via its Large Venture fund, with the participation of Elaia, its historical investor. The deeptech company, which specializes in the discovery of drug candidates, will be able to use this financing to develop its exclusive drug discovery platform and its portfolio of therapeutic assets, as well as to recruit. Maximilien Levesque, its co-founder, after post-doctorates in statistical physics and fluid mechanics at ENS Paris and the universities of Cambridge and Oxford, worked as a CNRS researcher at ENS. He developed a statistical algorithm inspired by quantum mechanics that could have applications in drug discovery and focused the work of his group on drug design. He joined forces with Emmanuelle Martiano, a strategy consulting engineer at the Boston Consulting Group who has worked for pharmaceutical and BtoB software companies, and the two of them created Aqemia in June 2019.

French start-up Iktos has announced a collaboration with Pfizer on the use of its artificial intelligence technology for drug design. This partnership comes in response to the considerable progress in the development of AI algorithms and computing power that has enabled the development of innovative approaches to small molecule drug design. Founded in 2016, Iktos develops generative AI technology in numerous collaborations with pharmaceutical and biotech companies. A fundamental aspect of the technology lies in the exploration of chemical space performed by generating compounds in silico under the constraints of the program's final objectives, rather than by screening compound libraries. As part of the collaboration, Pfizer has deployed Iktos' generative AI technology and is applying it to several small molecule research programs.