biostudies-literatureUnknownWang XJSchwartz/Reisman Collaborative Science ProgramHeritage Medical Research InstitutePew-Stewart Cancer Scholars programIsraeli Council for Higher Education (CHE) via the Weizmann Data Science Research CenterGordon and Betty Moore FoundationNational Institutes of HealthHHMI Freeman Hrabowski Scholar ProgramRita Allen FoundationSusan E Riley FoundationShurl and Kay Curci FoundationSchmidt Academy for Software EngineeringAbisch-Frenkel foundationEnoch foundation research fundSharon Levine FoundationEuropean Research CouncilRising Tide foundationIsrael Science FoundationNIH HHSNIGMS NIH HHSWe present DeepCell Types, a novel approach to cell phenotyping for spatial proteomics that addresses the challenge of generalization across diverse datasets with varying marker panels collected across different platforms. Our approach utilizes a transformer with channel-wise attention to create a language-informed vision model; this model's semantic understanding of the underlying marker panel enables it to learn from and adapt to heterogeneous datasets. Leveraging a curated, diverse dataset named Expanded TissueNet with cell type labels spanning the literature and the NIH Human BioMolecular Atlas Program (HuBMAP) consortium, our model demonstrates robust performance across various cell types, tissues, and imaging modalities. Comprehensive benchmarking shows superior accuracy and generalibioRxiv : the preprint server for biology2024.11.02.621624https://www.ebi.ac.uk/biostudies/studies/S-EPMC11601246biostudies-literatureGeneralized cell phenotyping for spatial proteomics with language-informed vision models.PMC11601246948811OT2OD033759DP2-GM149556OT2 OD033759DP2 GM149556OT2 OD033756OT2OD0337562481/203830/21Abt MBussi YPradhan EVan Valen DBorner KYu KWang XJBrown CKeren LBarnowski RLi SIqbal ARDilip RJain YYue YfalseGeneralized cell phenotyping for spatial proteomics with language-informed vision models.We present DeepCell Types, a novel approach to cell phenotyping for spatial proteomics that addresses the challenge of generalization across diverse datasets with varying marker panels collected across different platforms. Our approach utilizes a transformer with channel-wise attention to create a language-informed vision model; this model's semantic understanding of the underlying marker panel enables it to learn from and adapt to heterogeneous datasets. Leveraging a curated, diverse dataset named Expanded TissueNet with cell type labels spanning the literature and the NIH Human BioMolecular Atlas Program (HuBMAP) consortium, our model demonstrates robust performance across various cell types, tissues, and imaging modalities. Comprehensive benchmarking shows superior accuracy and generali2025-01-01T00:00:00Z2025 Aug2026-05-23T03:24:58.365Z2025-04-03T23:55:48.37ZS-EPMC116012463960565110.1101/2024.11.02.621624