A Coordinated, Self-Serve Path for Biospecimen Research at University of Virginia
Executive Summary
- Researchers requesting tissue or blood samples from UVA's Biorepository and Tissue Research Facility (BTRF) worked mostly through emails. Without a shared format or a direct line into the specimen system, fulfilling and tracking requests was time intensive.
- Manifold ingested more than 15,000 consented patient biospecimen records with automated monthly updates and linked them to the UVA Cancer Registry, so specimen availability and clinical criteria live in one place.
- Agents enabled researchers to query that linked data in plain language, identify a cohort with matching specimens, and send it straight to a shared workbench where BTRF staff review and fulfill the request.
- The BTRF team received a structured and auditable request instead of an email thread, so requests are easier to follow, and staff spend less time interpreting raw data.
Coordinating specimen requests in email chains took time and multiple back-and-forth
When a researcher needed tissue or blood from the UVA Biorepository and Tissue Research Facility (BTRF), the request usually started as an email. Without a shared format or direct access to the specimen system, moving a request forward could take some back-and-forth, and its status wasn't always easy to track. Also, because clinical criteria lived in one system and specimen availability in another, determining the availability of matched specimens often meant multi-touch coordination across teams and systems.
Unify the data, ask in plain language, and hand teams a request ready to fulfill
The work started by building on the existing foundation. The 15,000+ consented biospecimen records, previously siloed across separate systems, is ingested into Manifold and linked to the Cancer Registry, already on Manifold, with no need to rebuild the foundation. The records are kept current with automated monthly updates.
On top of the unified data, the platform applies fine-grained, role-based access control. What each person can see and query is based on the scope of their role and their approved level of access, keeping sensitive patient data governed and secure.
Unified, the rich data becomes available to an AI agent. A researcher asks a question in plain language, for instance, "find breast cancer patients with pre-treatment biopsies available," and the AI agent returns a cohort with matching specimens, checking clinical criteria and specimen availability simultaneously.
From there the cohort goes straight to a shared workbench, where BTRF staff can review and fulfill the request. An email chain is transformed into a structured handoff.
A single coordinated path from question to fulfilled request with every request traceable end-to-end
Unified data across silos
Information that lived in separate systems is now coordinated in one place. The 15,000+ consented biospecimen records were ingested into Manifold with automated monthly updates and linked to the Cancer Registry, so specimen and clinical data can be queried together.
A clean handoff between researchers and the biorepository team
Every request becomes a structured, auditable record the moment its made. A purpose-built biospecimen AI skill produces a clean, single joined table, so staff can act on a request without interpreting raw data. Requests, the data behind them, and the resulting cohorts are all accounted for in one place, making them easier to track than the earlier email-based process.
Less manual operational burden
A researcher describes what they need in plain language, the AI agent queries specimen availability alongside clinical criteria in a single interaction and surfaces the matching cohort with specimen availability. The handoff to BTRF is automated with a structured, auditable request in the workbench. One interaction now covers what used to take multiple systems, team members, and several rounds of email.
A capability the teams never had before
Manifold becomes the intake and coordination layer for specimen requests. Through a structured, self-serve path, researchers can confirm matching specimens exist before they ask. They can ask questions in plain language instead of manually interpreting raw data, and BTRF can fulfill requests from a structured queue instead of an inbox.
What's next
The shared workbench allows both researchers and the BTRF team to export the data they need, gated by role-based access controls, and layer a compute environment, JupyterLab notebooks and other specialized tools, with both low-code and high-code tools to match the analysis each researcher wants to run.
