LTMG

LTMG models transcriptional expression states in single-cell RNA-seq (scRNA-seq) data using a left-truncated mixture Gaussian framework to infer multimodal gene expression distributions and account for zero and low expression values.

Key Features:

• Modeling Transcriptional Expression States: Models kinetic relationships among transcriptional regulatory inputs, mRNA metabolism, and mRNA abundance to represent gene expression dynamics in single cells.
• Handling Zero and Low Expressions: Treats zero and low expression values as left-truncated data to improve inference of expression states.
• Expression Multi-Modalities Inference: Infers multi-modal gene expression distributions to capture cellular heterogeneity.
• Goodness of Fit: Demonstrates superior goodness of fit on extensive scRNA-seq datasets relative to three other state-of-the-art models.
• Biological Assumptions and Validation: Assumptions about low non-zero expressions are biologically rational and the multimodality setting is validated against independent experimental datasets.
• Differential Gene Expression Testing: Includes a differential gene expression test with higher sensitivity and specificity compared to five other popular methods.
• Co-Regulation Module Identification: Identifies co-regulation modules corresponding to perturbed transcriptional regulations.

Scientific Applications:

• Single-Cell RNA-Seq Data Analysis: Applied to scRNA-seq datasets to characterize cellular diversity and transcriptional regulation.
• Gene Expression State Inference: Captures multi-modal expression states to resolve functional heterogeneity within cell populations.
• Differential Expression Analysis: Used to identify differentially expressed genes with enhanced sensitivity and specificity across conditions or treatments.
• Co-Regulation Network Exploration: Enables detection of co-regulation modules for studies of transcriptional regulatory mechanisms.

Methodology:

LTMG applies a left-truncated mixture Gaussian model leveraging kinetic relationships between transcriptional regulatory inputs and mRNA metabolism/abundance and accounts for dropout events and low expression levels.