.png)
BIODEFENSE
BIODEFENSE AT THE SPEED OF EMERGING THREATS
Identifying neurotoxic risk and prioritize countermeasures quickly with scalable in vivo screening.
We combine zebrafish behavioral phenotyping, automated computer vision, and predictive models to deliver decision-ready leads in weeks, not years.
Chemical neurotoxic agents of concern can drive acute and long-lasting CNS effects, but traditional countermeasure discovery is slow, low-throughput, and costly. Lunai Bioworks provides an AI-based zebrafish platform that rapidly profiles neuroactive effects, classifies risk, and identifies candidate countermeasures at scale. We focus on detecting neurotoxic compounds and preventing downstream use through screening, prioritization, and technological safety filters.
HIGH-CONCERN AGENT SCREENING - RAPID NEUROTOXIC TRIAGE
WHAT WE DID:
- We built a scalable behavioral screening workflow in zebrafish embryos, recorded with a proprietary light stimulus battery, and analyzed via automated computer vision to extract activity traces.
OUTCOMES:
- A low-cost, rapid screening assay that classifies compounds as potentially neurotoxic by comparing new profiles to known behavioral signatures.
IMPLICATIONS:
- Faster, earlier identification of neurotoxic risk enables focused follow-up, reduces time spent on low-value candidates, and supports rapid response in time-sensitive scenarios.
COUNTERMEASURE DISCOVERY - HIGH-THROUGHPUT IN VIVO SCREENING
WHAT WE DO:
Use the same in vivo platform to counter-screen structurally diverse compound libraries against selected neurotoxic exposures and prioritize hits that reduce adverse behavioral effects, with confirmation across doses and additional agents.
OUTCOMES:
High-throughput screening of structurally diverse compound libraries with prioritized hits on week-scale timelines, supporting rapid iteration and rescreening as new agents come into scope.
IMPACT:
High-throughput, physiology-relevant discovery increases the probability of finding viable countermeasures and de-risks downstream validation by starting from a whole-organism signal.
LLM SAFETY LAYER - BIOLOGICAL RISK INTELLIGENCE FOR GENERATIVE AI
CURRENT WORK:
- Training a chemical transformer model on high-content zebrafish phenotyping data capturing neuroactive and systemic responses, then use that model as a safety layer within AI workflows to flag or block outputs predicted to drive neurotoxic or biosecurity-relevant phenotypes.
ANTICIPATED OUTCOMES:
- A biologically informed screening and governance capability that can support safety stress-testing, alignment, and safer deployment of chemistry- and biology-adjacent AI systems.
IMPACT:
- Organizations can reduce the risk of downstream misuse by adding in vivo grounded safety intelligence to generative tools before candidates are acted upon.
.avif)
