Open-source biomedical research software

NosoGraph

Disease Intelligence. Connected.

Open-source research software for connecting disease knowledge, evidence, and provenance across biomedical sources.

PUBLIC_ALPHA · RESEARCH USE ONLY · NO HOSTED DEMO YET

Illustrative NosoGraph evidence path A disease connects to phenotypes, genes, pathways, interventions, and claims. Claims connect to supporting or contradictory evidence, studies, and source provenance. Disease Phenotype Gene Pathway Variant Intervention Claim Evidence Evidence Study Source

Select a labeled entity to trace relationships.

Illustrative model, not a live query. NosoGraph keeps the claim, evidence direction, study context, and source provenance visible for inspection.

The problem

Biomedical evidence is connected in the literature, but scattered in practice.

Diseases, phenotypes, genes, pathways, trials, publications, and therapies live across different resources. Finding a relationship is only the beginning; researchers also need its context and lineage.

  • Distributed knowledge Ontologies, genetics resources, pathway databases, trials, and literature each expose a partial view.
  • Hard-to-trace claims A result without its underlying evidence is difficult to review or reproduce.
  • Uneven coverage A registry entry is not the same thing as a deeply curated disease module.
  • Uncertain context Species, study design, origin, and missing metadata change how evidence should be read.

The NosoGraph layer

Turn fragmented sources into inspectable relationships.

NosoGraph is a computational layer across upstream resources—not a replacement for them. It normalizes records, creates typed claims, attaches evidence, and preserves provenance for downstream inspection.

SourcesMONDO, HPO, PubMed, trials, genetics, pathways, and drug data.
NormalizeResolve identifiers and preserve source context.
ClaimsRepresent deterministic biomedical assertions.
EvidenceSeparate supports, contradicts, and inconclusive signals.
ProvenanceKeep snapshot, import, fingerprint, and source lineage visible.
MONDOHPO / HPOAPubMedClinicalTrials.govOpen TargetsGWAS Catalog

Evidence-first by design

See why information is present—not only what was generated.

Evidence records support claims; they are not automatic proof. Associations are not causation. Conflicting evidence can coexist, and unknown metadata remains unknown rather than becoming a confident score.

Read the Evidence Explorer guide
  1. Disease context
  2. Typed claim
  3. Supporting / contradictory evidence
  4. Study and source metadata
  5. Provenance and snapshot

The Evidence Explorer is a read-only public-alpha surface. Quality dimensions such as species, study design, origin, and human review describe context; sparse metadata is shown as unknown.

Real paths, not dead-end cards

Researcher path

From disease question to inspectable evidence.

  • Explore a disease and its connected entities
  • Inspect claims and evidence directions
  • Trace provenance to source snapshots
  • Compare relationships with explicit missingness
See a research example →
Developer path

From architecture to contribution.

  • Understand the disease and biomedical store model
  • Run the CLI, API, dashboard, and tests locally
  • Extend a source adapter or disease module
  • Validate the change and open a contribution
Read the contribution path →
PUBLIC_ALPHAResearch software under active development. Snapshot: v0.1.0 · 2026-08-22.

What exists today—and what does not.

Maturity labels are part of the product. They tell you how to interpret a surface before you use it.

CapabilityCurrent stateWhat that means
Unified nosograph CLIStableTask-oriented commands for exploration, validation, sources, analysis, and serving.
FastAPI API + dashboardBetaWorking local research interface with async jobs and documented API surfaces.
Evidence ExplorerPublic alpha · included in v0.1.0Read-only claim → evidence → provenance → source workflow.
Evidence WorkspaceBetaMulti-source evidence assembled into claims and ranked hypotheses.
NosoGraph CompareExperimentalInitial multidimensional comparison with explicit missingness.
Public hosted demoPlannedNot deployed; local Docker evaluation is the current route.
Optional LLM enrichmentExperimentalNot required for core deterministic extraction and evidence workflows.
FHIR / OMOP / PhenopacketsNot implementedFuture interoperability work, not current capability.
10,407registry modules
88strict L2-validated
6reference modules
8CI-validated modules
40+analysis pipelines
2,391offline tests in v0.1.0 suite

Snapshot source: public-status.yaml. Registry breadth is not curation depth; most registry modules are scaffolds.

Scientific and technical practice

Trust comes from inspectability.

Source transparency

Upstream resources, integration state, and data terms are documented in a source matrix.

Provenance

Source snapshots, import paths, filters, fingerprints, and retrieval context can travel with an output.

Conservative quality

Species, study design, origin, and human review are contextual dimensions; missing values stay unknown.

Reproducibility

CLI validation, fixture-backed workflows, tests, and explicit maturity make limits visible.

Open implementation · open questions

Read the code. Improve the model.

Add a source, deepen a disease module, improve the evidence workflow, or ask a research question. NosoGraph is built in public and labeled honestly.