UNIFAN

UNIFAN integrates unsupervised clustering with functional annotation of single-cell RNA-sequencing (scRNA-Seq) data using neural networks and cell-specific gene-set activity scores to improve cell-type assignment.

Key Features:

• Simultaneous clustering and annotation: Combines clustering and annotation by leveraging low-dimensional gene representations and cell-specific gene set activity scores.
• Neural network approach: Employs an unsupervised neural network to model scRNA-Seq data complexity and noise for improved cell-type assignment.
• Utilization of known gene sets: Uses predefined gene sets, including pathway gene sets, to guide clustering and to annotate biological processes associated with each cluster.
• Performance on benchmark datasets: Demonstrates superior performance relative to existing methods on human and mouse scRNA-Seq datasets across multiple organs by incorporating gene-set knowledge.

Scientific Applications:

• Cellular heterogeneity analysis: Provides evidence-based cluster annotations from scRNA-Seq data to dissect cellular heterogeneity.
• Developmental biology: Enables identification and annotation of cell types and states relevant to developmental processes.
• Immunology and disease pathology: Supports precise cell-type assignment and functional interpretation in immunological studies and disease-related tissue profiling.

Methodology:

UNIFAN uses an unsupervised neural network that integrates low-dimensional gene representations with cell-specific gene set activity scores derived from predefined gene sets to perform simultaneous clustering and functional annotation of scRNA-Seq data.