A latent factor framework to organize regulatory and metabolic programs inferred from scRNA-seq

Area of Science:

• Computational Biology
• Systems Biology
• Genomics

Background:

• Single-cell RNA sequencing (scRNA-seq) offers high-resolution transcriptional heterogeneity but limited insight into regulatory and metabolic processes.
• Computational methods infer transcription factor (TF) activity and metabolic features from RNA, providing complementary functional views.

Purpose of the Study:

• To develop a computational framework for jointly modeling multiple transcriptome-derived functional projections.
• To organize inferred regulatory and metabolic programs into an interpretable latent space for enhanced cellular state characterization.

Main Methods:

• Utilized a latent factor organizational strategy to jointly model gene expression, TF regulon activity, metabolite-level features, and predicted metabolic fluxes from scRNA-seq data.
• Applied the framework to a breast cancer cell line dataset.

Main Results:

• The latent space organized distinct inferred programs into coordinated axes of variation, guided by regulatory and metabolic constraints.
• Identified proliferative, oxidative-metabolic, and stress-associated functional programs in breast cancer cells, offering deeper insights than RNA-only analyses.
• Demonstrated that inferred regulatory and metabolic programs can be structured into an interpretable latent representation.

Conclusions:

• Joint modeling of transcriptome-derived functional projections provides a more coherent functional characterization of cellular states.
• The proposed framework facilitates functional interpretation beyond gene expression alone by integrating diverse biological information.