Post 16 CompanyValency bond

RefinedScience Partners with Valency to Accelerate Drug Development Research

Valency MCP Bond expands the reach of RefinedScience's AI co-scientist across tens of millions of research papers, helping scientists uncover new connections and sharpen the hypotheses that drive drug development.

Today we’re announcing a partnership between Valency and RefinedScience, a biology-first precision medicine company building the biological foundation for better drug development.

RefinedScience brings together primary human tissue, single-cell molecular data, longitudinal clinical data, and laboratory validation to help drug developers make decisions grounded in real human biology. Now, with Valency MCP Bond connected to their internal AI co-scientist, the team can identify current research that supports, challenges, or helps clarify scientific hypotheses, with underlying sources they can open and assess.

Evidence you can trust

Drug development is unforgiving. In one analysis of more than 400,000 clinical trial entries, only 3.4% of oncology drug programs that entered clinical trials went on to approval.1 The decisions that shape a program are only as strong as the evidence beneath them.

Meanwhile, the research landscape is changing under everyone’s feet. Nearly 90% of open-access biomedical papers published by the end of 2025 show signs of large language model (LLM)-assisted writing.2 And the AI tools scientists lean on have their own blind spots. Large language models are trained on older science, can miss recent findings, and will sometimes hand back a confident answer with no source, or a citation that doesn’t exist.

For a team whose work depends on getting the biology right, “sounds plausible” isn’t good enough. RefinedScience needs to trust the research it relies on, and to connect its findings to sources a scientist can open and check. Bond gives the RefinedScience team a wider field of discovery, bringing emerging research into their investigations of human disease. With unexpected connections to explore and original sources at hand, the team can challenge assumptions and identify promising directions for the next experiment.

Grounded in the literature

Built by researchers, for researchers, Valency MCP Bond is a research intelligence layer that gives AI systems access to more than 45 million papers and preprints across PubMed, arXiv, bioRxiv, medRxiv, and other sources. Its capabilities include semantic search across research content, author disambiguation, and citation and collaboration mapping, helping researchers find relevant work and understand how discoveries connect. New research is added daily, keeping the evidence current.3

Through the Model Context Protocol (MCP), Bond connects directly to existing AI assistants and agents without requiring researchers to adopt a separate platform. RefinedScience has integrated Bond as the literature-access component of its internal AI co-scientist, combining current scientific literature with its proprietary biological and clinical datasets and other research sources.

The result is an AI workflow grounded in verifiable scientific evidence. Rather than relying solely on what a model learned during training, RefinedScience’s scientists can access current research, trace findings to their sources, and make more informed decisions.

For the RefinedScience team, that means:

  • Verifiable sources: Find research to help evaluate and contextualize scientific findings, with papers scientists can open, review, and cite.
  • Current research: Access newly published papers and preprints as they’re added to Bond’s continuously updated corpus.
  • Seamless integration: Access Bond directly within existing AI workflows, without learning a separate platform.
  • Connected discoveries: Explore citation networks to understand how findings build on and influence other research.

From answers to evidence

For RefinedScience, the impact extends beyond finding relevant papers. By integrating Bond into its AI co-scientist, the team can connect scientific findings to verifiable evidence and surface research that general-purpose AI tools might otherwise miss.

“Valency Bond brings current scientific literature into our research workflow. Alongside our biological and clinical data, that context helps us evaluate hypotheses, sharpen our research focus, and decide which questions to investigate next.”

— Grant Weller, PhD, Chief Technology Officer, RefinedScience

That’s exactly the outcome we had in mind when we built Bond: in a field where decisions carry real weight, make the evidence behind an insight current and easy to verify.

“The next breakthrough in drug development could begin with a connection between discoveries that have never been considered together. RefinedScience brings deep evidence from human biology, and Bond helps its scientists connect that evidence with discoveries across scientific literature. That creates opportunities to challenge established thinking and take new questions into the lab.”

— Josh Bloom, CEO and Co-founder, Valency

Get started with Valency MCP Bond

Bond is an LLM-agnostic tool for researchers that empowers users to verify any claim against its source with a single click. Sign up for Bond here.

To talk with our team about a similar setup for your organization, contact us at sales@valency.io.

To learn more about RefinedScience and partnership opportunities, visit RefinedScience.com or email partnerships@refinedscience.com.


  1. Wong, C. H., Siah, K. W. & Lo, A. W. (2019). “Estimation of clinical trial success rates and related parameters.” Biostatistics, 20(2), 273–286. Based on 406,038 trial entries for 21,143+ compounds, 2000–2015; the 3.4% oncology figure is the aggregate success rate from the data sample. ↩︎

  2. Holzwarth, L., González-Márquez, R. & Kobak, D. (2026). “Most biomedical publications show signs of LLM-assisted writing.” arXiv:2608.10715 (preprint). Estimates that 89% of PubMed Central open-access papers show excess LLM-associated vocabulary by the end of 2025. ↩︎

  3. Ram, K. (2026). “Valency Bond Is Live for All Users.” Valency Blog, 1 October 2026. Corpus size and daily refresh. ↩︎