Predictive modeling of molecular activity underlying physical cell-cell interactions.

Area of Science:

• Immunology
• Computational Biology
• Genomics

Background:

• Cell-cell interactions are crucial for tissue function and disease development.
• Labeling immune partnerships by sortagging intercellular contacts (LIPSTIC) quantifies direct physical interactions.
• Single-cell RNA sequencing (scRNA-seq) profiles intracellular transcriptomes.

Purpose of the Study:

• To present Gloss (group lasso on scRNA-seq), a predictive modeling framework.
• To systematically link gene and pathway activity to LIPSTIC-measured interaction strength.
• To provide a computational framework for analyzing LIPSTIC+scRNA-seq data.

Main Methods:

• Developed the Gloss predictive modeling framework.
• Applied group lasso regression to link gene expression with interaction data.
• Validated Gloss against correlation-based and standard regression approaches.

Main Results:

• Gloss outperforms existing methods in linking gene activity to cell interaction strength.
• Identified molecular features of myeloid-T cell interactions in anti-Ctla4 immunotherapy.
• Described interactions between T cell subpopulations during viral infection.

Conclusions:

• Gloss is an interpretable and effective framework for analyzing LIPSTIC+scRNA-seq data.
• The framework prioritizes genes and pathways driving cellular communication.
• Gloss advances the study of cell-cell interactions in health and disease.