Metavert
Aitiome grades whether an environmental chemical has mechanistically-supported links to Parkinson's or Alzheimer's.
Core trick
Scores chemicals by curated evidence along reconstructed MIE-to-AO pathways, deliberately excluding bioactivity signals from the grade.
Description
Aitiome grades whether an environmental chemical has mechanistically-supported links to Parkinson's or Alzheimer's. Give it a chemical and it reconstructs the OECD-endorsed causal pathway (MIE→KE→AO), grounds each edge in queryable evidence, and grades it on curated evidence — never bioactivity, with calibrated confidence. On a curated adversarial benchmark it recovers 13/13 PD and 12/12 AD known neurotoxicants and rejects 6/6 mitochondria-active decoys with zero errors. The key finding, computed live from our own data: bioactivity is anti-diagnostic — every activity signal separating real neurotoxicants from decoys is at or below chance, while the curated rule is perfect. It ships the discovery limits as an honest map, adds a wet-lab candidate-triage queue, and exposes the whole engine over MCP for agents. It also includes a methods study: adversarial RLM surfaced ~10× more counter-evidence than RAG.
README excerpt
# Aitiome **An honest mechanistic-reasoning engine for the environmental exposome of neurodegeneration.** `Go` · `React 18 + TypeScript + Vite` · `Three.js / GLSL` · `MCP` · `Claude Opus 4.8` · `MIT` · *Built with Claude: Life Sciences*  > Submitted by **Jon Radoff** for the **Built with Claude: Life Sciences** hackathon. Give Aitiome a chemical and it reconstructs the OECD-endorsed causal pathway — molecular initiating event (MIE) → key events (KE) → adverse outcome (AO) — to a Parkinson's or Alzheimer's hallmark, grades it on **curated evidence (never bioactivity)**, and rates its confidence. It proves itself two ways at once: it **recovers** the known neurotoxicants on the endorsed scaffold, and it **refuses to be fooled** by bioactive compounds that are not neurotoxic. It is **not** a system that claims to discover novel neurodegeneration-causing chemicals. It is transparent about exactly where AI-driven discovery is, and is not, possible on this chemical class — and it ships that honesty as a feature. --- ## ▶ Try it live - **App → https://aitiome.fly.dev** - **Presentation deck → https://aitiome.fly.dev/presentation.pdf** Click a **known neurotoxicant** to watch the endorsed cascade reconstruct and ignite the vulnerable dopaminergic-neuron terminal. Click a **bioactive decoy** to see it withheld and rejected on independent lines of evidence. --- ### A methods result: how should Claude synthesize scientific evidence?