NAMO: a unified ontology for New Approach Methodology metadata

The problem NAMO is solving

New Approach Methodologies — the collective term for non-animal alternatives in biomedical research — have matured quickly. Brain organoids, liver-on-chip devices, QSAR toxicity predictors: each generates data with real scientific value. But each also tends to come with its own terminology, its own reporting conventions, and its own database that does not talk to the others.

OrganoidDB catalogs transcriptomics for 16,000+ organoid samples. MPS-DB covers pharmacokinetic endpoints from microphysiological systems. Cell Model Passports focuses on cancer model genomics. None of these can straightforwardly be queried together, and none was designed to accommodate the others' data types. The result is a situation familiar from traditional animal research: detailed data locked in formats that prevent cross-study comparison and systematic regulatory evaluation.

The passage of the FDA Modernization Act 2.0 in 2022 — which eliminated mandatory animal testing requirements for drug development — has made this fragmentation more urgent. Regulatory acceptance of NAMs requires structured, comparable evidence. The data infrastructure for that does not yet exist in a unified form. NAMO is an attempt to provide it.

What NAMO is and how it is structured

NAMO (New Approach Methodology Ontology) is developed by the Monarch Initiative and built on LinkML — a schema language that generates Python dataclasses, OWL ontology files, JSON Schema, and GraphQL schemas from a single YAML source definition. The key practical advantage of this approach is consistency across formats without manual synchronisation.

The ontology organises all model systems under a single abstract root class (ModelSystem), which branches into AnimalModel (retained for comparison purposes) and NAMModel. The NAMModel branch then divides into three major categories reflecting genuinely different experimental logics:

• Cellular systems — 2D cultures, 3D cultures, spheroids, organoids, co-cultures. The Organoid class alone adds 23 properties specific to self-organising structures: differentiation methods, maturation timelines, branching morphogenesis indicators.
• Microphysiological systems — organ-on-chip and tissue-on-chip devices. This category captures microfluidic channel architecture, membrane specifications, flow parameters (perfusion rate, shear stress), and mechanical stimulation parameters (e.g., breathing simulation frequency for lung chips).
• In silico models — QSAR, physiologically-based pharmacokinetic (PBPK) models, machine learning classifiers, digital twins, and metabolic models. Algorithm specifications, training data characteristics, and performance metrics are formalised as structured slots.

Current v0.1.0 statistics: 48 classes, 156 properties, 12 enumerations, 6 integrated external ontologies. The schema is published under a CC0 license and is also browsable on BioPortal.

Ontology integration and interoperability

Rather than redefining terms already established elsewhere, NAMO uses LinkML's reachable_from mechanism to reference branches of external ontologies and validate that annotations stay within those branches. A hepatocyte is CL:0000182; liver tissue is UBERON:0002107; cystic fibrosis is MONDO:0009563. These annotations carry the full semantic weight of their source ontologies — including hierarchical relationships and cross-references — without NAMO having to maintain those definitions itself.

The six integrated ontologies are: UBERON (13,000+ anatomical structures), Cell Ontology (2,400+ cell types), ChEBI (140,000+ chemical entities), NCBITaxon (species), OBI (experimental design concepts), and ECO (evidence and conclusion types). This means a NAMO-annotated organoid record is, in principle, directly queryable alongside records from any database that uses the same ontology terms — which is the practical foundation of semantic interoperability.

The concordance framework: structured validation evidence

One of the more interesting design decisions in NAMO is how it handles validation. Most NAM reporting treats validation as a binary or as unstructured text. NAMO introduces a StructuredConcordanceResult class that decomposes validation into five dimensions, each with its own quantitative metrics and methodological metadata:

• Molecular similarity — gene expression correlation coefficients, protein abundance patterns, metabolite profiles, with specific genes or metabolites listed and confidence intervals recorded
• Pathway concordance — pathway activation scores (GSEA, PROGENy), directional consistency of regulation
• Phenotype overlap — quantified using standardised phenotype ontologies, with severity and penetrance metrics
• Functional parity — physiological readouts such as TEER for barrier integrity, beat frequency for cardiac models
• Statistical measures — confidence intervals, p-values, effect sizes for all the above

This is a meaningful step beyond current practice. The manuscript notes that published organoid validations typically report concordance for only 10–50 marker genes, while comprehensive transcriptomic comparisons remain rare — a gap made visible precisely because the schema provides a place to record both what was assessed and what was not. Having structured absence of evidence is as informative as having evidence.

Practical relevance for preclinical researchers

NAMO is at an early stage (v0.1.0, preprint manuscript, 15 GitHub stars at the time of writing). It is not yet a deployed database standard. But it is worth tracking for several reasons:

First, the Monarch Initiative has a credible track record in biomedical ontology infrastructure — this is not a one-off academic exercise. Second, the FDA Modernization Act 2.0 pressure is real and creates an institutional need for exactly the kind of structured validation evidence NAMO formalises. Third, the LinkML-based multi-format generation approach solves a genuine practical problem: the same schema produces OWL for semantic reasoning, JSON Schema for API validation, and Python classes for programmatic access — without having to maintain three separate artefacts.

For teams working with organoids or organ-on-chip alongside traditional animal or HCM data, NAMO is worth watching as a potential bridge: a common metadata layer that could make multi-modal, cross-platform comparisons tractable without starting from scratch. The ontology is freely available at github.com/monarch-initiative/namo and documentation at monarch-initiative.github.io/namo.