CellTICS

CellTICS performs interpretable cell-type identification and pathway characterization from single-cell RNA sequencing (scRNA-seq) data.

Key Features:

• Biological Interpretability: Focuses on marker genes with cell-type-specific expression to link predictions to biologically meaningful features.
• Hierarchical Pathway-Based Construction: Constructs a neural network using a hierarchy of biological pathways to map pathway contributions to cell-type definitions.
• Multi-Predictive-Layer Strategy: Employs multiple predictive layers to resolve broad cell types and finer sub-cellular classifications.
• Pathway Identification and Characterization: Identifies and characterizes pathways that define specific cell types or are altered under conditions such as disease or aging by leveraging the network's nonlinear responses.
• Expression Stochasticity: Detects pathways where expression stochasticity—variability in gene expression levels—contributes to cell-type identity.

Scientific Applications:

• Cellular Heterogeneity Analysis: Maps pathways and marker genes that underlie cellular heterogeneity in scRNA-seq datasets.
• Developmental Biology: Dissects pathway dynamics and cell-type transitions during development using scRNA-seq data.
• Pathology and Disease Mechanism: Identifies pathways and cell-type-specific alterations associated with disease states.
• Aging Studies: Detects pathway changes and expression variability associated with aging.
• Regenerative Medicine: Characterizes cell-type identities and pathways relevant to regeneration and cell-fate engineering.

Methodology:

Uses a nonlinear neural network architecture constructed from a hierarchy of biological pathways with a multi-predictive-layer strategy to prioritize marker genes and identify pathway-level contributions to cell-type predictions.