scGTM

scGTM models gene expression trends along cell pseudotime to identify interpretable monotone, hill-shaped, and valley-shaped patterns in single-cell RNA-seq count data.

Key Features:

• Interpretable Trends: Captures monotone, hill-shaped, and valley-shaped gene expression patterns while maintaining interpretability comparable to simpler models and flexibility comparable to nonparametric approaches.
• Biologically Informative Parameters: Uses parameterization intended to be biologically interpretable and informative about gene expression trends.
• Flexible Distribution Accommodation: Accommodates common distributions for modeling gene expression counts in single-cell RNA-seq data.
• Optimization via Particle Swarm Algorithm: Employs particle swarm optimization to find constrained maximum likelihood estimates for parameter inference.

Scientific Applications:

• Dynamic Biological Processes: Analyze gene expression trends along pseudotime to investigate dynamic processes at the single-cell level.
• Developmental Pathways: Identify molecular signatures and timing associated with developmental pathways.
• Cellular Differentiation: Characterize gene expression changes underlying cellular differentiation.

Methodology:

Models the statistical relationship between pseudotime and gene expression counts and uses particle swarm optimization to identify constrained maximum likelihood parameter estimates.