Technical optimization of spatially resolved single-cell transcriptomic datasets to study clinical liver disease.

Single cell and spatially resolved 'omic' techniques have enabled deep characterization of clinical pathologies that remain poorly understood, providing unprecedented insights into molecular mechanisms of disease. However, transcriptomic platforms are costly, limiting sample size, which increases the possibility of pre-analytical variables such as tissue processing and storage procedures impacting RNA quality and downstream analyses. Furthermore, spatial transcriptomics have not yet reached single cell resolution, leading to the development of multiple deconvolution methods to predict individual cell types within each transcriptome 'spot' on tissue sections. In this study, we performed spatial transcriptomics and single nucleus RNA sequencing (snRNASeq) on matched specimens from patients with either histologically normal or advanced fibrosis to establish important aspects of tissue handling, data processing, and downstream analyses of biobanked liver samples. We observed that tissue preservation technique impacts transcriptomic data, especially in fibrotic liver. Deconvolution of the spatial transcriptome using paired snRNASeq data generated a spatially resolved, single cell dataset with 24 unique liver cell phenotypes. We determined that cell-cell interactions predicted using ligand-receptor analysis of snRNASeq data poorly correlated with celullar relationships identified using spatial transcriptomics. Our study provides a framework for generating spatially resolved, single cell datasets to study gene expression and cell-cell interactions in biobanked clinical samples with advanced liver disease.