BenchSci decodes disease biology with Neo4j and Google Cloud to accelerate drug discovery

BenchSci transforms fragmented biomedical evidence into a graph-native model of disease biology, empowering 9 of the top 10 pharmaceutical companies to de-risk and accelerate discovery. BenchSci’s ASCEND platform combines large language models with the Biological Evidence Knowledge Graph (BEKG), an experimentally grounded knowledge system that integrates open-access literature, closed-access publications, and proprietary pharmaceutical datasets. Its LENS extraction engine uses Google Cloud Vertex AI and Gemini models to interpret scientific papers with contextual awareness and validate claims against associated imagery before materializing structured assertions into Neo4j.

Organization
BenchSci
Industry
Pharma
Location
Canada
Published
March 2026

Reported outcomes

Strategic outcomes

Better decisions & insightTraceable evidence-backed inferenceNew product / capabilityBuilt a graph-native evidence layerNew product / capabilityEnabled multimodal scientific extractionCustomer experience & trustReduced manual literature review

Primary read

Use case focus

Showing 3 of 4

  • 1Knowledge Graph
  • 2Retrieval Augmented Generation
  • 3Multimodal AI
  • Biomedical evidence is fragmented across millions of publications, experimental contexts, model systems, and proprietary datasets.
  • Disease biology is networked and multi-hop reasoning was brittle in relational databases, creating performance and provenance issues.
  • Scientists needed traceable, evidence-backed inference to reduce late-stage drug failures caused by misunderstood biology.
  • Built the Biological Evidence Knowledge Graph on Neo4j as a graph-native evidence layer for disease biology.
  • Used Google Cloud Vertex AI and Gemini models in the LENS engine for multimodal extraction from papers and images.
  • Used BigQuery for large-scale analytics and retrieval-augmented generation over subgraphs to ground generative responses in experimentally validated relationships.
  • ASCEND is deployed by 9 of the top 10 global pharmaceutical companies, including Sanofi and Merck.
  • ASCEND is used by more than 4,500 research institutions worldwide.
  • The platform reduces manual literature review and helps scientists identify mechanistic gaps and conflicting evidence earlier in discovery.
Architecture

BenchSci built the Biological Evidence Knowledge Graph (BEKG) on Neo4j and uses Google Cloud Vertex AI with Gemini in its LENS engine to extract experimentally grounded assertions from scientific papers and imagery. Structured outputs are materialized into Neo4j, while BigQuery supports large-scale analytics. Scientists query the system through a retrieval-augmented workflow that retrieves relevant subgraphs before generation, preserving provenance and traceability.

Implementation partners3
Sources & evidence1
Groundedness: 5/5Type: Customer StoryPublished: Mar 3, 2026Publisher: Neo4jEvidence: PrimaryConfidence: High

AI-generated summary. Verify important details with the linked sources before relying on this case.

Explore related AI use cases

Was this useful?

Similar cases