NeuroLab AI

Summary

CNS discovery is slowed less by missing data than by fragmented data. Chemistry, target pharmacology, anatomy, circuits, imaging, cognition and clinical evidence sit in separate tools, so mechanism reasoning is assembled by hand and rarely auditable. NeuroLab is an auditable neurosymbolic platform that connects this evidence across scales and returns inspectable mechanism hypotheses — exposing the route used, supporting and contradicting evidence, applicability limits, and reasons to abstain.

Problem

Fewer than 10% of CNS candidates entering first-in-human trials reach approval; the clinical approval rate for CNS compounds is roughly 6.2% versus 13.3% outside CNS. Evidence linking a compound to receptor, cell and layer, circuit, region, cognition and clinical observation is scattered across disconnected sources. Most tools return one opaque score without exposing the route, its applicability domain, or the missing transition — so a supported inference looks identical to a plausible one.

Impact

One governed evidence substrate with three explicit routes — a graph predictor for known compounds, a chemistry-to-receptor retriever for novel structures, and a deterministic symbolic composer — plus provenance, abstention contracts and a 392-parcel 3D Brain Explorer. Discovery teams can triage compounds and mechanisms, locate the evidence gap blocking an inference, compare alternatives and design discriminating experiments. Next milestone: a frozen, dual-disjoint benchmark of the full chain.