Detecting Data Embedding Spatial Patterns and Identifying Biomarkers with BioRSP

Abstract

AbstractMotivationHigh-dimensional data embedding has become an essential tool in bioinformatics, particularly for single-cell data analysis, as it aids in characterizing cells with complex gene expression patterns based on projected low-dimensional data point spatial distributions. However, traditional methods often fall short in distinguishing subtle changes within cell populations. To address this, we developed BioRSP (BiologicalRadarScanningPlots), an innovative open-source software designed to enhance the characterization of single-cell gene expression patterns by simulating radar beam scanning across defined clusters from a fixed coordinate, referred to as the “vantage point.”ResultsBioRSP enables users to analyze the distribution of each gene within and across cell clusters using a quantitative measure known as the RSP plot. These plots categorize genes and spatial embedding patterns into four distinct groups based on combinations of gene coverage (high/low) and RSP values (high/low), reflecting spatial distribution regularities in the embedding space. This novel approach allows BioRSP to distinguish subtle differences between background cells and those expressing specific markers above a preset threshold, significantly improving upon traditional gene-to-gene correlation analyses. A case study utilizing a neonatal mouse heart tissue dataset from the Mouse Cell Atlas demonstrates BioRSP’s capability to identify spatially distinct and functionally significant gene expression patterns.AvailabilityThe Python package and its comprehensive documentation are publicly available athttps://github.com/cytronicoder/biorsp.Contactcytronicoder@gmail.com