Non-negative Independent Factor Analysis disentangles discrete and continuous sources of variation in scRNA-seq data

Area of Science:

• Computational biology
• Genomics
• Bioinformatics

Background:

• Single-cell RNA sequencing (scRNA-seq) is crucial for understanding cellular heterogeneity.
• Dimensionality reduction (DR) is essential for scRNA-seq data analysis but often lacks biological interpretability.
• Interpretable DR is needed to link reduced dimensions to biological variables like cell type or pathway activity.

Purpose of the Study:

• To develop a novel probabilistic model for scRNA-seq data analysis.
• To create a DR method that enhances biological interpretability by disentangling cell identity and pathway activity.
• To provide a unified framework for analyzing complex single-cell data.

Main Methods:

• Introduced Non-negative Independent Factor Analysis (NIFA), a probabilistic factor analysis model.
• NIFA simultaneously models uni- and multi-modal latent factors.
• Applied NIFA to various scRNA-seq datasets, including an immunotherapy dataset.

Main Results:

• NIFA effectively disentangles discrete cell-type identity and continuous pathway activity.
• NIFA-derived factors outperformed existing methods (ICA, PCA, NMF, scCoGAPS) in separating biological variation.
• NIFA reproduced, refined, and discovered new cell states in an immunotherapy dataset.

Conclusions:

• NIFA offers a powerful and interpretable approach for scRNA-seq data analysis.
• The model facilitates the discovery of novel biological insights and clinically relevant cell states.
• NIFA provides a unified framework for dissecting cellular heterogeneity.