Detecting autism at birth and identifying children likely to respond to treatment: long out of reach, these prospects are becoming concrete thanks to artificial intelligence. Driven by my work, two major French advances, both world firsts, could transform the early care and treatment of autism spectrum disorders (ASD), which affect approximately 1 in 127 children worldwide. While diagnosis still occurs on average between the ages of 4 and 6, AI is paving the way for earlier, more precise medicine that is better adapted to the diversity of individual profiles.
Detecting Autism in the First Days of Life
To address this major public health challenge, the Pelargos project, led by B&A Biomedical, relies on artificial intelligence to analyze data from routine maternity monitoring and the first days of life, in order to identify newborns at risk of autism spectrum disorder at a very early stage.
A first pilot study conducted in 2021 validated its relevance, and a national validation phase is underway in five university hospitals (CHUs) in France to confirm these results on a large scale, with a target of 2,000 children included. Ultimately, the Pelargos diagnostic tool could allow children to be referred to appropriate interventions from the earliest years of life, during the crucial window of brain plasticity. The stakes are high: to sustainably improve developmental trajectories, autonomy, and quality of life for affected children.
Beyond diagnosis, Pelargos thus paves the way for true predictive medicine applied to neurodevelopmental disorders. France could become one of the first countries to enable early identification of autism risk from birth.
Treatments: Artificial Intelligence Sheds New Light on Drug Development
At the same time, another major breakthrough, led by Neurochlore, a company dedicated to therapeutic research in ASD, is transforming the search for treatments in autism. While no drug treatment is currently approved for core symptoms, artificial intelligence now makes it possible to revisit clinical trials from a new perspective.
Despite seven successful international Phase 2 trials, a large-scale Phase 3 clinical trial conducted by Neurochlore in partnership with Servier, aimed at market approval, did not meet its primary endpoints.
Neurochlore, in collaboration with B&A Biomedical, conducted a retrospective analysis of this Phase 3 trial. By reanalyzing all clinical data using machine learning tools, researchers demonstrated that the treatment used – bumetanide – is in fact effective in 30% to 40% of the children included, who show a positive response.
Previously undetectable due to the heterogeneity of patient profiles, this efficacy now makes it possible to identify responder subgroups and specific clinical profiles. A new optimized clinical trial, based on these criteria, could lead to the first treatment targeting a significant proportion of children with autism.
This approach marks a major turning point: the objective is no longer to find a universal treatment, but to tailor care to each individual child.
Towards a More Personalized Approach to Autism
These advances pave the way for a profound transformation in the management of autism spectrum disorders. Until now, this field has been marked by late diagnosis, broad approaches, and often inconclusive clinical trials.
This work now outlines a new model based on earlier detection, early-life interventions, and treatments tailored to each child’s profile. While nearly 2% of children are affected, diagnostic pathways remain long and challenging for families, with significant human, educational, and economic impacts.
In this context, these advances offer concrete perspectives for all stakeholders – families, healthcare professionals, and public decision-makers – and illustrate a new dynamic: that of earlier, more personalized medicine, better adapted to the diversity of ASD.
