← Back to Home

From Neuron Imaging to Drug Discovery: Recursion and Genentech Advance Their First Neuroscience Target

A target identified through large-scale cell experiments and AI analysis has passed three stages of validation and entered small-molecule drug discovery. This marks a step forward for the collaboration, but the target’s identity and detailed data remain undisclosed, and demonstrating therapeutic efficacy is still a long way off.

By SURL BioNews

In drug discovery for neurological diseases, identifying a disease-related gene is only the starting point; the real challenge is determining whether altering it can improve cellular disease phenotypes. On August 5, 2026, Recursion announced that the first neuroscience target identified through its collaboration with Genentech had passed experimental validation and entered a joint small-molecule discovery program. The next step will be to find molecules that act on this target.

This development also appeared in an overview of investor discussions compiled by Quiver Quantitative on September 1. However, that report primarily summarized market views on the R&D pipeline and added no new experimental evidence. The specific biological progress remains grounded in Recursion’s announcement: after reviewing the validation results, Genentech exercised its validated target option under the collaboration agreement.

To establish a foundation for identifying targets, the team reportedly produced more than one trillion neurons differentiated from human induced pluripotent stem cells, conducted knockout experiments covering more than 17,000 genes, and obtained more than 30 million images of neuronal cells. The AI models were tasked with comparing cellular changes caused by these genetic perturbations to identify genes linked to known disease mechanisms that had not yet been sufficiently explored.

After the models generated leads, the team tested them in stages. First, it confirmed whether the pathways involving the target matched the predictions. Next, it tested whether manipulating the target could alter the health, metabolism, and electrophysiological function of disease-relevant human neurons. Finally, it examined whether disease phenotypes changed as a result. The company said the target passed all three stages before being cleared to advance.

This takes the program across a threshold from computational prediction to experimental support. However, the public announcement has not disclosed the target’s name, the corresponding disease, effect sizes, or complete experimental data, leaving readers unable to independently assess the strength of the validation. Large numbers of cells and images indicate the scale of discovery; they do not constitute evidence of efficacy in patients, nor can these numbers alone establish whether AI has improved the success rate of R&D.

Next, Recursion plans to use its chemistry platform to design and optimize molecules, gradually turning the biological lead into potential drug candidates. Even if the target is validated, whether molecules can act effectively, have sufficient selectivity, and achieve acceptable safety will require separate validation. This collaboration offers an example of early target discovery; whether it can produce a drug suitable for human trials will be the more practical test ahead.

References

  1. Quiver Quantitative
  2. Recursion Pharmaceuticals