Every recommendation now stands on one graph.

We've fused ninety-plus sources into a single evidence-typed knowledge graph: the central dogma materialised end to end, gene-level coverage across 20k+ organisms, and layers of binding, regulation, clinical, and target-discovery evidence the co-scientist reasons over. It's what turns a recommendation into something you can trace, not just trust. Counts below are from the current graph snapshot (September 2026); an entity is one canonicalised node and a relation is one directed, evidence-typed edge.

We show you what doesn't work, too.

Most graphs only remember hits. This one keeps curated negatives right alongside the positives, so a "no" carries as much weight as a "yes." None of it is sampled or inferred.

Negatives are curated, experimentally-confirmed non-interactions and inactive screening outcomes. An inactive result records inactivity under the reported assay conditions; it does not establish non-binding in general. Positive evidence spans clinical genetics, drug–target pharmacology, precision-medicine associations, and disease ontology.

Twenty-six entity types, one graph.

Genes dominate: 123M+ across vertebrates, plants, fungi, invertebrates, protists, and ~20k bacterial genomes. But that's just the start. Transcripts, exons, chemicals, diseases, pathways, and 20 more entity types all live in the same graph, ready to traverse together. Bars are log-scaled to keep every type visible.

From 7 species to 20k+ organisms.

The central dogma is now materialized end to end (gene → transcript → exon → coding → protein) for 20k+ organisms, with genomic-neighbor edges giving every gene its operon/synteny context. Function annotation alone now spans 13k+ species.

Genome assemblies across the tree of life were parsed into the graph, with genes created wherever a clade isn't already covered. Function annotations carry evidence-tagged relationships (enables, involved_in, located_in) across 13k+ species.

Find everything like it, through one shared node.

Every protein connects to proxy nodes for what it shares: its family, the ligands it binds, its modifications, its functional keywords, its domains. Two proteins in the same family, or binding the same cofactor, meet at that shared node: two edges, one traversal, no custom query required.

From binding to bedside, modeled.

The molecular, clinical, and regulatory core below sits alongside the tree-of-life, shared-attribute, and RNA layers above; 80 predicates in all, each one a relationship the co-scientist can reason across. Bars are scaled within each layer.

Real edges. Not illustrations.

A live edge pulled from the graph for every predicate below, with entity types, evidence, and sign: actual rows, not mockups.