NIFA

NIFA performs Non-negative Independent Factor Analysis to decompose single-cell RNA sequencing (scRNA-seq) data into uni-modal and multi-modal latent factors that separate discrete cell-type identities and continuous pathway activities for improved interpretability of dimensionality reduction.

Key Features:

• Probabilistic Modeling Framework: Implements a probabilistic factor analysis model that integrates assumptions about data structure to produce more interpretable dimensionality reduction outputs.
• Uni- and Multi-modal Latent Factor Analysis: Models uni-modal and multi-modal latent factors simultaneously to separate discrete cell-type identities from continuous pathway activities.
• Enhanced Biological Interpretation: Isolates biological variables such as cell-type identity and pathway activity to clarify inter-cellular heterogeneity.
• Comparative Performance: Demonstrates superior performance in disentangling biological variation relative to Independent Component Analysis (ICA), Principal Component Analysis (PCA), Non-negative Matrix Factorization (NMF), and scCoGAPS in datasets with known cell-type identities.
• Application to Diverse Datasets: Has been applied across multiple datasets, including immunotherapy studies, to replicate and refine findings and identify clinically relevant cell states.

Scientific Applications:

• Cell-type Identification: Isolates discrete cell-type identities for classification and characterization of cell populations in complex tissues.
• Pathway Activity Analysis: Models continuous pathway activities to enable exploration of gene expression dynamics linked to biological pathways.
• Clinical Research (Immunotherapy): Identifies novel cell states with potential clinical significance in immunotherapy datasets.

Methodology:

Non-negative Independent Factor Analysis implemented within a probabilistic factor analysis framework that models uni-modal and multi-modal latent factors for dimensionality reduction, with comparative evaluation against ICA, PCA, NMF, and scCoGAPS.