{"entity": "publication", "iuid": "a3408d291a75474e99d7f92ea9fb850c", "timestamp": "2026-08-20T20:46:18.939Z", "links": {"self": {"href": "https://publications-affiliated.scilifelab.se/publication/a3408d291a75474e99d7f92ea9fb850c.json"}, "display": {"href": "https://publications-affiliated.scilifelab.se/publication/a3408d291a75474e99d7f92ea9fb850c"}}, "title": "INLAomics for Scalable and Interpretable Spatial Multiomic Data Integration.", "authors": [{"family": "Arnroth", "given": "Lukas", "initials": "L", "orcid": "0000-0002-8567-5116", "researcher": {"href": "https://publications-affiliated.scilifelab.se/researcher/2dc2c8f75f5742c6b165d10aba742391.json"}}, {"family": "Vickovic", "given": "Sanja", "initials": "S", "orcid": "0000-0003-0985-9885", "researcher": {"href": "https://publications-affiliated.scilifelab.se/researcher/2d4188f01ed74ba499ef845e0ad9fe97.json"}}], "type": "journal article", "published": "2025-05-08", "journal": {"title": "bioRxiv", "issn": "2692-8205", "issn-l": null}, "abstract": "Integrating spatial transcriptomics with antibody-based proteomics enables the investigation of biological regulation within intact tissue architecture. However, current approaches for spatial multi-omics integration often depend on dimensionality reduction or autoencoders, which disregard spatial context, limit interpretability, and face challenges with scalability. To address these limitations, we developed INLAomics, a multivariate hierarchical Bayesian framework that models protein abundance in tissue sections by leveraging histological features and latent spatial factors inferred from spatial transcriptomics data. INLAomics supports two key applications: (1) identifying spatial gene co-expression programs to build interpretable gene-protein networks, and (2) predicting spatial protein expression in tissues lacking proteomics measurements. Applied across diverse datasets, INLAomics reveals previously unrecognized gene-protein associations and achieves substantial improvements in protein prediction accuracy over models that treat each modality independently. The framework is both computationally efficient and biologically interpretable, offering a scalable solution for integrative analysis of spatial multi-omics data.", "doi": "10.1101/2025.05.02.651831", "pmid": "40654897", "labels": [], "xrefs": [{"db": "pmc", "key": "PMC12247973"}, {"db": "pii", "key": "2025.05.02.651831"}], "notes": [], "created": "2026-08-20T11:02:42.259Z", "modified": "2026-08-20T11:03:32.261Z"}