MANU's research lab · Core Disease Intelligence v1.0

A research brief that shows its sources, not just its conclusions.

Give MedSolve Labs a disease name and it returns a structured brief — overview, mechanisms, standard of care — built from live literature and database queries, with every claim traced to a PMID, an Open Targets entry, or a UniProt record. When two sources disagree, it says so instead of picking one quietly.

See how grounding works What this is (and isn't)

Citation typesPMID · Open Targets · UniProt

Grounding audit98.5% acc.

Research time saved~14.5 hrs / brief

Statusv1.0 in review

DISEASE BRIEF Alzheimer's disease · generated 03/09/2026 3 sections

Professional research use only. MedSolve Labs synthesizes research literature. It does not provide medical advice and must not be used for diagnosis or treatment. The brief above is a real, live-generated example; disease names and citation IDs are real, not staged.

Grounding

The anatomy of one brief

A generated paragraph is only as trustworthy as what it's built from. Every section of a MedSolve Labs brief carries its own consistency check, separate from the brief's overall accuracy score.

Inline citationsPMID · OT · UniProtLiterature, target–disease association, and protein-structure sources are cited inline, not summarised into a generic reference list.

Consistency checkMulti-sourceEach section is checked against more than one source and labelled accordingly, rather than treated as equally certain by default.

Issues this brief1 flaggedThe standard-of-care section rests on a single source here and is marked as such, instead of being smoothed over.

Grounding audit98.5% acc.An automated audit rechecks generated claims against their cited sources after the brief is produced.

Modules

Beyond a single disease brief

The disease brief is the most visible surface, but the same grounding discipline runs through the rest of the lab — from target identification to a structured look at why a past trial didn't work.

Targets

Core

Gene and protein targets implicated in a disease, with the association evidence behind each one.

Organisms

Reference data

Model-organism data relevant to a target or pathway, alongside the human evidence.

Molecules

Reference data

Compound-level data tied to a target, for the molecules already implicated in it.

Trial Failure Autopsy

Analysis

A structured breakdown of why a past clinical trial failed, not just the fact that it did.

Audit

Grounding

The accuracy log behind every generated brief, reviewable rather than taken on faith.

Session Q&A

Follow-up questions, same rule

A brief raises follow-up questions. The session Q&A answers them — constrained strictly to the references already cited in the brief, rather than reaching for outside knowledge.

GROUNDED SESSION Q&A · SOURCE GROUNDED

"What's the evidence behind the receptor-targeting mechanism?"

Per UniProt:P12345, structural mapping shows abnormal folding and active-site mutations altering receptor binding — consistent with the pathway association in OT:ENSG00000130600. Both sources agree on the mechanism; the open question is patient-specific efficacy, which is why Standard of Care carries the issue flag above.

Demo · a fixed example question and answer, shown to illustrate the constraint — not a live model call.

Scope

What this is, and what it isn't

MedSolve Labs is MANU's own research-lab initiative, separate from the RealActivity collaboration — a place to test ideas that might later become their own project, not a single fixed product.

Built and real

Core Disease Intelligence v1.0

Disease-brief generation against live literature and database queries, with inline citations, per-section consistency checks, and a grounding audit — built, and currently under review before merging into the main line.

PMID, Open Targets, and UniProt citations, live
Multi-source consistency checks per section
Trial Failure Autopsy and grounding audit

Not this

Not a diagnostic or treatment tool

MedSolve Labs synthesizes research literature for professional research use. It is not medical advice, not a diagnostic device, and every output needs expert review before it informs a real decision.

No fixed roadmap — ideas here may spin out or be dropped
Not tied to a single deliverable, unlike MANU's other projects

Get in touch

Interested in this project?

MedSolve Labs is the most open-ended of MANU's current projects — a good one to talk about if you have a research workflow it might fit.

Talk to us about MedSolve Labs

Or ask what v1.0's review is currently focused on.

Start a conversation Read the documentation