Pathway metrics accurately stratify T cells to their cells states

Abstract

Pathway analysis is a powerful approach for elucidating insights from gene expression data and associating such changes with cellular phenotypes. The overarching objective of pathway research is to identify critical molecular drivers within a cellular context and uncover novel signaling networks from groups of relevant biomolecules. In this work, we present PathSingle, a Python-based pathway analysis tool tailored for single-cell data analysis. PathSingle employs a unique graph-based algorithm to enable the classification of diverse cellular states, such as T cell subtypes. Designed to be open-source, extensible, and computationally efficient, PathSingle is available at https://github.com/zurkin1/PathSingle under the MIT license. This tool provides researchers with a versatile framework for uncovering biologically meaningful insights from high-dimensional single-cell transcriptomics data, facilitating a deeper understanding of cellular regulation and function.

Keywords: Dimensionality reduction; Machine learning; Pathway analysis; RNA sequencing; Single-cell data; Systems biology.

Fig. 1

Gaussian scaling activity of two variables

Fig. 2

PathSingle pipeline

Fig. 3

Clustering UMAP results on PBMC dataset

Fig. 4

A UMAP representation of the pathways dimension reduction of single-cell expression data reduced to a map in which the different T cell states are of distinct spatial behavior

Fig. 5

Using only 3 features (pathways) to represent the data. The original 357 dimensions of the single-cell data was reduced using PathSingle metrics, to a map in which the different T cell states are of distinct spatial behavior. The image provides a stratified representation of T cells states