Researching biomedical data systems and publishing the tools openly for research and discovery.
Comprehensive, deterministic biomedical catalog — 52,000+ canonical reference pages on human genes, drugs, and diseases, built by mining the BioBTree knowledge graph. Every figure traces to primary sources. Designed for researchers and AI agents.
Unified knowledge graph built from 80+ biomedical datasets — genes, proteins, diseases, drugs, and ontologies, with the connections between them. Billions of evidence-backed edges link every entity. Query it directly over a REST API and MCP server, built for AI-assisted research.
An open system predicting human protein targets across ~30M SureChEMBL patent compounds — by chemical similarity to 1.25M known ChEMBL ligands. Two directions from one index: a compound’s likely targets, or a target’s patented chemistry with patent provenance. Grounded to stable identifiers via BioBTree.
A deterministic per-variant reference for germline human variants — 3.9M across 18,799 genes — built by mining the BioBTree knowledge graph. Each variant consolidates its ClinVar classification, gnomAD frequency, and calibrated effect predictors, flagging where the computational and clinical evidence disagree.
An open benchmark for anticipating which variants of uncertain significance (VUS) will be reclassified, and in which direction, while they are still uncertain. Built by temporal holdout over dated ClinVar releases — 580,386 variants scored on their later outcomes.
A content-neutral workflow coordinator where humans, autonomous AI agents, and deterministic compute tasks claim and work the same DAG as peers. Upstream results feed automatically into the tasks that depend on them, and every result is committed to git — so each run stays reproducible and auditable.