ShanYiran/TIST: TIST， Transcritpome and histopathological Image integrative analysis for Spatial Transcriptomics

RNA-seq based spatial transcriptomics is an emerging bio-technique to study in situ gene expression heterogeneities and patterns at whole-genome scale. Identification of spatial domain, which is a region with similar gene expression patterns, is one of the fundamental tasks for analyzing the spatial transcriptomic data. Except transcriptomic data, the technique usually generated histopathological images for the same tissue slide at the same time. Compared to the noisy and low-resolution spatial transcriptomic data, histopathological images show high spatial continuity and relatively low noise, and provide complementary macroscopic cellular information of tissue. Hence, we proposed a novel spatial domain identification algorithm called TIST(Transcritpome and histopathological Image integrative analysis for Spatial Transcriptomics), which can integrate both the transcriptomic data and the corresponding histopathological image information. TIST used Markov random field (MRF) to learn the macroscopic cellular features from histopathological images, and then it devised a random walk based strategy to integrate the extracted image features and the spatial transcriptomic data for domain segmentation. We tested TIST on diverse samples including brain, several types of tumors and kidney. Results show TIST can achieve better domain continuity by reducing the noises of individual tissue spots and preserve the sensitivity to identify small functional regions at the same time.

install.packages("remotes")
remotes::install_github("ShanYiran/TIST")