Leveraging spatial transcriptomics data to recover cell locations in single-cell RNA-seq with CeLEry.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) provides high-resolution cellular heterogeneity insights but lacks spatial context.
• Understanding cell location is crucial for interpreting biological processes in tissues.
• Existing scRNA-seq methods cannot determine the physical origins of cells.

Purpose of the Study:

• To develop a computational method for recovering spatial cell locations from scRNA-seq data.
• To integrate gene expression data with spatial transcriptomics information.
• To enhance the application of scRNA-seq in studying tissue architecture and disease.

Main Methods:

• Developed CeLEry (Cell Location recovEry), a supervised deep learning algorithm.
• Leveraged gene expression and spatial location relationships from spatial transcriptomics.
• Incorporated an optional variational autoencoder for data augmentation to improve robustness against noise.

Main Results:

• CeLEry successfully infers 2D cell locations and spatial domains from scRNA-seq data.
• The algorithm provides uncertainty estimates for the recovered spatial locations.
• Benchmarking across multiple datasets (brain, cancer) and spatial technologies (Visium, MERSCOPE, MERFISH, Xenium) confirmed CeLEry's reliability.

Conclusions:

• CeLEry effectively recovers spatial information for cells using scRNA-seq data.
• The method overcomes limitations of dissociated cell analysis by inferring spatial origins.
• CeLEry offers a robust approach for integrating scRNA-seq with spatial context, advancing tissue-level biological research.