Clustering and visualization of single-cell RNA-seq data using path metrics

Recent advances in single-cell technologies have enabled high-resolution characterization of tissue and cancer compositions. Although numerous tools for dimension reduction and clustering are available for single-cell data analyses, these methods often fail to simultaneously preserve local cluster structure and global data geometry. To address these challenges, we developed a novel analyses framework, Single-Cell Path Metrics Profiling (scPMP), using power-weighted path metrics, which measure distances between cells in a data-driven way. Unlike Euclidean distance and other commonly used distance metrics, path metrics are density sensitive and respect the underlying data geometry. By combining path metrics with multidimensional scaling, a low dimensional embedding of the data is obtained which preserves both the global data geometry and cluster structure. We evaluate the method both for clustering quality and geometric fidelity, and it outperforms current scRNAseq clustering algorithms on a wide range of benchmarking data sets.Author summary: Advancements in single-cell technologies with the ability to measure gene expression at the cellular level have provided unprecedented opportunity to investigate the cell type (T cells, B cells, etc) and cell state diversity (active T cells and exhausted T cells) within tissues and cancers. However, analyzing this complex high-dimensional data when the noise level is high requires sophisticated tools to effectively extract useful biological information and faithfully visualize the data in a low-dimensional space (2- or 3-D). Existing computational methods such as dimension reduction and clustering (group similar cells together) for single-cell data struggle to simultaneously preserve local group structure and global data geometry (developmental relationship between cell types). To tackle this problem, we’ve developed a new analysis framework called scPMP (Single-Cell Path Metrics Profiling) based on a unique approach to measure distances between cells which takes into account both the density of cells (common vs rare cell types) and the overall structure of the data. We have demonstrated the ability of scPMP to better preserve the natural grouping of cells and the relationships between different groups over existing methods in numerous real and simulated data sets. This improvement could lead to more accurate identification of cell types and states.