B-Lightning: using bait genes for marker gene hunting in single-cell data with complex heterogeneity

Area of Science:

• Single-cell genomics
• Computational biology
• Biostatistics

Background:

• Single-cell studies face challenges with multiple sources of heterogeneity (SOH) that can obscure important biological signals.
• Nuisance SOH, such as cell type or cell cycle phase, can confound the identification of the SOH of interest, impacting accurate cell subpopulation annotation.

Purpose of the Study:

• To develop B-Lightning, a novel and robust computational method for identifying marker genes and cell subpopulations related to a specific SOH.
• To isolate the SOH of interest from other confounding SOH in single-cell data.

Main Methods:

• B-Lightning employs an iterative approach to enrich reliable marker genes.
• It enhances the signals of the SOH of interest, effectively boosting their detectability.
• The method is validated through multiple numerical simulations and experimental studies.

Main Results:

• B-Lightning demonstrates superior sensitivity and robustness in marker gene identification compared to existing methods.
• It significantly improves the ability to differentiate cell subpopulations of interest from other heterogeneous groups.
• The method successfully identified novel markers for senescence, T-cell responses, Alzheimer's disease, and dendritic cell function in various disease contexts.

Conclusions:

• B-Lightning is a powerful tool for single-cell data analysis, especially in complex datasets with entangled SOH.
• It offers enhanced accuracy and power for identifying biologically relevant cell subpopulations and their markers.
• The method has broad applicability across diverse biological and medical research areas.