In the race to accelerate drug discovery through artificial intelligence, French startup Iktos has emerged as a pioneer in Europe, developing an integrated platform that combines generative AI, automated synthesis, and biological testing.
The company's 8-year journey reflects both the tremendous potential and the practical challenges of applying AI to one of the most complex and crucial tasks in medical science.
While the pace of discovery seems to have suddenly accelerated, Iktos CEO Yann Gaston-Mathé said that the industry's evolution has been a gradual buildup that is only now bursting into public awareness.
"It's not like 10 years ago there was nothing, and now there's AI," he said in an interview. "It's been a constant evolution since the early 80s."
Still, the company has shifted into higher gear over the past year following fundraising, an acquisition, and a more robust platform rollout. The challenge isn't just technological but cultural, requiring pharmaceutical companies to adapt their mindset and approaches to drug discovery.
Iktos now faces the challenge of transforming a traditionally manual, time-intensive process to leverage the potential of AI to reinvent this industry fully.
Deep Learning Innovation
The story of Iktos begins with a convergence of expertise in pharmaceutical research and emerging AI technologies. Gaston-Mathé, an engineering graduate from École Polytechnique, spent years in pharmaceutical R&D at Servier and Ipsen before transitioning to data science.
The company's genesis came when two data scientists – CSO Quentin Perron and CTO Nicolas Do Huu – approached him with a novel idea: using deep generative models – what would later be known as generative AI – to help medicinal chemists identify better drug candidates.
"They were really pioneers in the field," Gaston-Mathé recalled. "At the time, there was absolutely no example in the scientific literature of such works."
The fundamental challenge they sought to address was the near-infinite possibilities in chemical space when designing new molecules. Drug candidates need to meet multiple complex criteria: novelty for patenting, potency, selectivity, good ADME properties, and absence of toxicity.
