Abstract Biological techniques for spatially resolved transcriptomics (SRT) have advanced rapidly in both throughput and spatial resolution. This progress calls for efficient and scalable spatial dimension reduction methods capable of handling large-scale SRT data from multiple tissue sections. Here, we developed FAST, a fast and efficient generalized probabilistic factor analysis model for spatially aware dimension reduction. FAST simultaneously accounts for the count-based nature of SRT data and extracts low-dimensional representations across multiple sections, while preserving biological signals and incorporating spatial smoothness among neighboring locations. Unlike existing methods, FAST explicitly models count data across sections and leverages local spatial dependencies with scalable computational complexity. Using both simulated and real datasets, we demonstrated that embeddings estimated by FAST show improved correlation with annotated cell and domain types. Notably, FAST was the only method capable of analyzing a mouse embryo Stereo-seq dataset with 2.3 million spatial locations in just 2 hours. FAST also identified differential activity of immune-related transcription factors between tumor and non-tumor clusters and predicted the carcinogenesis factor CCNH as an upstream regulator of differentially expressed genes in a breast cancer Xenium dataset. FAST is available for non-commercial use at https://github.com/feiyoung/ProFAST.
Liu et al. (Sat,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: