LEGEND: Identifying Co-expressed Genes in Multimodal Transcriptomic Sequencing Data

Abstract

Identifying co-expressed genes across tissue domains and cell types is essential for revealing co-functional genes involved in biological or pathological processes. While both single-cell RNA-sequencing (scRNA-seq) and spatially-resolved transcriptomic (SRT) data offer insights into gene co-expression patterns, current methods typically utilize either data type alone, potentially diluting the co-functionality signals within co-expressed gene groups. To bridge this gap, we introduce LEGEND, a novel computational method that integrates scRNA-seq and SRT data for identifying groups of co-expressed genes at both cell type and tissue domain levels. LEGEND employs an innovative hierarchical clustering algorithm designed to maximize intra-cluster redundancy and inter-cluster complementarity, effectively capturing more nuanced patterns of gene co-expression and spatial coherence. Enrichment and cofunction analyses further showcase the biological relevance of these gene clusters, and their utilities in exploring context-specific novel gene functions. Notably, LEGEND can reveal shifts in gene-gene interactions under different conditions, furnishing insights for disease-associated gene crosstalk. Moreover, LEGEND can be utilized to enhance the annotation accuracy of both spatial spots in SRT and single-cells in scRNA-seq, and pioneers in identifying genes with designated spatial expression patterns. LEGEND is available at https://github.com/ToryDeng/LEGEND.