Inference of marker genes of subtle cell state changes via iLR: iterative logistic regression

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) analysis often faces challenges in identifying small, informative gene sets for subtle cell state differences.
• Existing methods may not effectively pinpoint key marker genes crucial for understanding disease or treatment effects.

Purpose of the Study:

• To develop and validate a novel method, iterative logistic regression (iLR), for identifying small, informative marker gene sets from scRNA-seq data.
• To enhance the interpretability and efficiency of marker gene selection in complex biological datasets.

Main Methods:

• Iterative logistic regression (iLR) was employed, incorporating Pareto front optimization to balance gene set size and classification performance.
• The iLR method was benchmarked on in silico datasets against state-of-the-art approaches for single-cell classification.
• iLR was applied to real-world datasets, including distinguishing neuronal subtypes in autism spectrum disorder and analyzing immunotherapeutic effects in tumor microenvironments.

Main Results:

• iLR demonstrated comparable performance to existing methods while utilizing a significantly smaller fraction of genes for single-cell classification.
• The method successfully identified disease-relevant genes for distinguishing neuronal subtypes in healthy versus autism spectrum disorder patients with high accuracy.
• iLR inferred informative genes that showed cross-species translational potential (mouse-to-human) in tumor microenvironment studies, predicting entinostat's effects on myeloid cell differentiation.

Conclusions:

• Iterative logistic regression (iLR) offers an effective approach for inferring interpretable transcriptional signatures from complex scRNA-seq data.
• The identified gene sets possess prognostic or therapeutic potential, facilitating deeper biological insights.
• iLR provides a valuable tool for advancing marker gene discovery in various biological and clinical research areas.