Public metabolomics repositories host thousands of studies, but inconsistent metadata structure, vocabulary variation, and repository-specific labels limit reliable search, comparison, and reuse both within and across repositories. The authors introduce HARMONY, an ontology-guided harmonization framework and web platform designed to reconcile these disparities and enable unified discovery of metabolomics studies across MetaboLights and Metabolomics Workbench.
The reported motivation is to close gaps introduced by raw-text matching and single-field lookups, which fail to retrieve related studies when terminology or metadata field layouts differ between repositories.
HARMONY is presented as a combined data-processing framework and user-facing web service. The system preserves original deposited terms as evidence while mapping repository-specific labels into shared query terms via ontologies. The platform aims to make study discovery traceable: users can see harmonized terms and the original metadata that supported mappings.
The design emphasises two complementary capabilities: broad extraction of candidate records from multiple metadata sources and ontology-based mapping to canonical terms. The authors also integrate machine learning encoders to support retrieval by mapping study metadata into shared latent representations of biological and analytical context.
HARMONY resolves eight biological and analytical metadata nodes at study level:
A ninth node, metabolite identity, maps metabolite entities to RefMet across both repositories, enabling metabolite-level queries that use a shared reference vocabulary. Importantly, original repository terms are retained as evidence for each harmonization decision, preserving provenance and supporting auditability.
HARMONY's workflow comprises two main stages. First, a multi-source extraction stage retrieves records that single-field or naïve lookups might miss. This step aggregates metadata from multiple repository fields and sources to build a comprehensive input for mapping.
Second, ontology mapping converts repository-specific labels into shared query terms drawn from controlled vocabularies. This mapping reduces the cross-repository retrieval gap by aligning semantically equivalent but textually different labels under common ontology terms. The platform records the mapping path and retains the deposited terms as evidence for downstream inspection.
Across the full corpus of studies in the two evaluated repositories, HARMONY increased cross-repository retrievability from 75.5% (raw matching) to 89.6% after harmonization. The harmonization process yielded thousands of study-node retrievals and reconnected studies that raw-text search would have left unreachable within their own repositories.
Coverage of harmonized metadata nodes was high: about 91% of Metabolomics Workbench studies and 85% of MetaboLights studies had at least six of the eight biological/analytical nodes harmonized. These figures indicate substantial alignment of study-level contextual metadata between repositories after ontology-guided processing.
The HARMONY web platform, available at https://omicsinharmony.in, supports the following user functions:
Search results and filters are backed by harmonized metadata and retain links to the original deposited terms so users can verify provenance and mapping evidence.
The platform incorporates machine learning encoders that convert study metadata into shared representations of biological and analytical context. These encoders support retrieval and comparison by providing a learned mapping between diverse metadata inputs and common conceptual spaces.
Traceability is emphasised throughout: harmonized nodes include the original repository labels as evidence, and harmonization decisions are presented so users can audit mappings between deposited terms and ontology terms.
The manuscript reporting HARMONY is a preprint posted July 27, 2026, and has not undergone peer review. The web platform is publicly accessible at the project URL provided in the article. The authors disclose competing interests: PW holds equity in Clarity Bio Systems India Pvt. Ltd., and KJ holds equity in RadAI Pvt. Ltd. The preprint is distributed under a CC-BY-NC 4.0 International license.
HARMONY is positioned by the authors as a community-facing tool to improve discoverability, comparison, and reuse of public metabolomics studies by harmonizing metadata and metabolite identities across major repositories.