VISTA Uncovers Missing Gene Expression and Spatial-induced Information for Spatial Transcriptomic Data Analysis.

Area of Science:

• Spatial Transcriptomics
• Computational Biology
• Genomics

Background:

• Understanding cell activities in spatial context is crucial for deciphering tissue organization and function.
• Single-cell RNA sequencing (scRNA-seq) provides comprehensive gene expression but lacks spatial information.
• Subcellular spatial transcriptomics (SST) offers high-resolution spatial data but profiles a limited number of genes.

Purpose of the Study:

• To develop a novel computational approach, VISTA, for predicting unobserved gene expression in spatial transcriptomics data.
• To address the gene profiling limitations of current SST technologies.
• To enable a wider range of downstream analyses by imputing missing gene expression data.

Main Methods:

• VISTA utilizes variational inference and geometric deep learning to jointly model scRNA-seq and SST data.
• The method incorporates uncertainty quantification to provide reliable predictions.
• The approach is designed for efficient analysis of large-scale SST datasets, considering time and memory constraints.

Main Results:

• VISTA demonstrated superior performance in gene expression imputation across four diverse SST datasets.
• The method showed satisfactory time efficiency and memory consumption for large-scale data analysis.
• Imputed data enabled novel downstream applications, including detection of spatially variable genes and ligand-receptor interactions.

Conclusions:

• VISTA effectively predicts gene expression in spatial transcriptomics data, bridging the gap between scRNA-seq and SST.
• The approach facilitates advanced analyses such as spatial RNA velocity inference and in-silico perturbation studies.
• VISTA enhances the decomposition of spatial and intrinsic cellular variations, improving biological insights from spatial transcriptomics.