21 March 2024 | Siyu He, Yinuo Jin, Achille Nazaret, Lingting Shi, Xueer Chen, Sham Rampersaud, Bahawar S. Dhillon, Izabella Valdez, Lauren E. Friend, Joy Linyue Fan, Cameron Y. Park, Rachel L. Mintz, Yeh-Hsing Lao, David Carrera, Kaylee W. Fang, Kaleem Mehdi, Madeline Rohde, José L. McFalice-Figueroa, David Blei, Kam W. Leong, Alexander Y. Rudensky, George Plitas, Elham Azizi
Starfysh is a computational toolbox that integrates spatial transcriptomic (ST) data and histology images to characterize tissue-specific cell states and spatial hubs in complex tissues. It uses a deep generative model that incorporates archetypal analysis and known cell type markers to improve the characterization of spatial dynamics in tissues without requiring single-cell reference data. Starfysh enhances the integration of multiple tissue samples, enabling the identification of shared or sample-specific niches and underlying cell-cell crosstalk. The tool has been applied to breast cancer subtypes, revealing spatial heterogeneity in tumor-infiltrating immune cells and the role of spatial orientation in immune differentiation. It also identified an immunosuppressive niche in metaplastic breast cancer (MBC) characterized by Treg cell infiltration and metabolic reprogramming. Starfysh's ability to integrate and analyze complex ST data makes it a powerful tool for studying spatial heterogeneity in health and disease.Starfysh is a computational toolbox that integrates spatial transcriptomic (ST) data and histology images to characterize tissue-specific cell states and spatial hubs in complex tissues. It uses a deep generative model that incorporates archetypal analysis and known cell type markers to improve the characterization of spatial dynamics in tissues without requiring single-cell reference data. Starfysh enhances the integration of multiple tissue samples, enabling the identification of shared or sample-specific niches and underlying cell-cell crosstalk. The tool has been applied to breast cancer subtypes, revealing spatial heterogeneity in tumor-infiltrating immune cells and the role of spatial orientation in immune differentiation. It also identified an immunosuppressive niche in metaplastic breast cancer (MBC) characterized by Treg cell infiltration and metabolic reprogramming. Starfysh's ability to integrate and analyze complex ST data makes it a powerful tool for studying spatial heterogeneity in health and disease.