We are developing therapies for disorders driven by dysfunctional neural circuits. To develop these therapies, we are building an agentic discovery platform that learns from experiments in our own in vivo lab.
Our agents use CircuitATLAS to connect disease phenotypes with the neural circuits, cell types and molecular mechanisms, identify therapeutic targets and design experiments to test them. Our cross-species models learn from large-scale neural activity data to identify conserved circuit states and help predict how an intervention’s effects may translate from preclinical models to humans.
At our Data Foundry in Manila, our team of RAs test these hypotheses in vitro and in vivo. The results refine our models and guide the next set of experiments. This lab-in-the-loop approach allows our agents to develop and test ideas against experimental evidence throughout discovery.
We are starting with epilepsy, with the goal of developing first-in-class therapies across a full spectrum of disorders driven by dysfunctional brain activity.
Publications:
CircuitATLAS (https://arxiv.org/abs/2610.09643). Target discovery platform which combines mechanistic reasoning with an in vivo lab-in-the-loop. AI agents identify therapeutic targets and design experiments that our research teams execute in the lab. The resulting biological data feeds back into the discovery process to iteratively refine hypotheses and validate new interventions.
Cross species representation learning (https://arxiv.org/abs/2610.11222). Aligning mouse and human neural activity to predict clinical drug efficacy from preclinical data.
