Inspiration

Talking with industry experts, we found inspiration in SOLIX Technologies' approach of treatment discovery through researching unexplored connections between drugs and diseases. Our app plans to automate this process in a way that leads to safe, effective, and efficient trials.

What it does

Our app uses live public databases from ChemBL, Open Targets, and ClinicalTrials.gov to gather data on existing trials and treatment/delivery methods, building an interactive knowledge graph to visualize how different drugs are related. An AI agent then uses this data to summarize the findings of current trials and make recommendations on what experiments to run next, based on links that have yet to be investigated.

Challenges we ran into

When building the first prototype, we initially neglected to consider how patent law and certain safety considerations would affect drug trial proposals. After getting feedback, however, our app now takes both of these factors into account.

How it's built

Frontend: Next.js, React, TypeScript and Tailwind CSS, with Cytoscape.js for the interactive knowledge graph and Recharts for potency charts
Backend: Next.js API routes for benchmarking, trials, graph building/expansion, reports and Q&A, validated with Zod
AI: Claude Opus 5.5 via the Anthropic SDK, returning structured JSON so every claim carries source IDs
Data: queried live per search (nothing precomputed), with disk and memory caching for speed
Trust layer: server-side citation validation, evidence-type labels (measured / computed / clinical / AI summary), and a demo mode with cached responses

What we learned

As relative newcomers to the industry, we learned about some of the challenges facing pharma today, especially related to drug discovery and fiscal efficiency when deciding what trials to run.