Browse BinfoNet evidence

Biomedical evidence intelligence

BinfoNet

BinfoNet turns biological questions into inspectable publication evidence.

Input

Biological question

  • genes or a user-supplied protein module
  • species when specified
  • disease or biological process
  • the question to investigate
Output

Traceable evidence package

  • ranked publication titles and abstracts
  • explicit evidence selection
  • abstract-reported claims linked to PMIDs
  • validation results and run provenance

BinfoNet is actively being developed and evaluated.

How BinfoNet works

Biological questionGenes · species · disease · process
Graph retrievalGene-linked publications
Ranked evidenceRelevant titles and abstracts
Cited answerAbstract-reported claims with PMIDs
Evidence infrastructure
20.6M
publications
353.9K
species-specific genes
117.2K
context entities
Human · mouse · rat
species
Graph relations
18.7M
STRING protein associations
474.1M
citation links
214.8M
publication–entity mentions
Explore the graph-based evidence workflow →

How BinfoNet is evaluated

Retrieval, synthesis, citations, and automated evaluation are measured separately so a failure can be located rather than hidden inside one end-to-end score.

Explore Biomedical GenAI Evaluation →

Evidence in practice

The public evidence case traces a biological question from gene-linked publication candidates through evidence selection and structured synthesis to citation auditing.

Inspect the evidence case →

Current status

Development status

Demonstrated
  • gene–publication candidate retrieval
  • gene-coverage and semantic ranking
  • saved evidence selection and structured synthesis
  • PMID validation and Phoenix/OpenTelemetry traces
In development or evaluation
  • STRING-based protein-module expansion
  • citation-neighborhood and publication-community retrieval
  • broader expert relevance and grounding review
  • automated-judge calibration and biomedical-use readiness