CEP-IP: An Explainable Framework for Cell Subpopulation Identification in Single-cell Transcriptomics

Abstract

Single-cell RNA sequencing (scRNA-seq) frameworks lack explainable approaches for identifying cell subpopulations harboring strong pairwise monotonic gene-module relationships between a gene of interest (GOI) and its co-expressed genes. CEP-IP is introduced as a novel explainable machine learning framework to address this gap. In the primary dataset, TRPM4 served as the GOI and its co-expressed ribosomal genes (Ribo) were identified via Spearman-Kendall dual-filter (i.e., dual-filtered gene, DFG). Generalized additive modeling quantified TRPM4-Ribo relationship strength via deviance explained (DE), which was then mapped to individual cells via CEP classification to identify top-ranked explanatory power (TREP) cells. TRPM4-Ribo transcriptional space was then stratified into pre-IP and post-IP regions using inflection point (IP) analysis, producing four subpopulations per patient for pathway analysis. TRPM4-Ribo modeling outperformed alternative gene set modules (FDR<0.05). In each prostate cancer (PCa) patient, CEP-IP yielded four cell subpopulations, where pre-IP TREP cells showed enrichment of immune-related processes, and post-IP TREP cells were enriched for ribosomal, translation, and cell adhesion pathways. Validation was performed in the Allen middle temporal gyrus (MTG) and Neftel glioblastoma (GBM) datasets. In the MTG dataset (CARM1P1-DFG module), post-IP TREP cells showed enrichment of neuron projection ontologies. In the GBM dataset, FOXM1 was the sole GOI yielding mesenchymal-state DFGs, with FOXM1-DFG post-IP TREP cells enriched for cell division and microtubule pathways; 3D trajectory analysis demonstrated continuous trajectories of TREP cells that were obscured in 2D embeddings. CEP-IP identifies biologically distinct cell subpopulations in three independent scRNA-seq datasets, and it may be applicable to other pairwise GOI-DFG modules in single-cell transcriptomics.