CSHMM-TF

CSHMM-TF integrates transcription factor information into Continuous-State Hidden Markov Models to model temporal gene expression and regulatory dynamics in time series single-cell RNA-seq (scRNA-Seq) data.

Key Features:

• Integration with TF Information: Incorporates TF–target relationships during model construction to improve cell state reconstruction and temporal state assignment from scRNA-Seq.
• Probabilistic Modeling: Uses Continuous-State Hidden Markov Models (CSHMM) to represent continuous transitions in gene expression across time.
• TF Assignment and Activation Order: Assigns TFs to specific activation points and models their effect on emission probabilities of cells at later time points to determine regulatory order.
• Validation Across Datasets: Validated on multiple mouse and human scRNA-Seq datasets, identifying known and novel TFs with activation timing concordant with expression data and prior knowledge.
• Combinatorial Predictions: Predicts combinatorial TF interactions that are corroborated by established biological interactions.

Scientific Applications:

• Developmental Biology: Elucidating the order and timing of TF activation for developmental pathways and lineage specification.
• Cancer Research: Studying tumor progression and heterogeneity by identifying key regulatory events and potential therapeutic targets.
• Stem Cell Research: Modeling regulatory networks driving stem cell differentiation through dynamic gene expression analysis.

Methodology:

Implements Continuous-State Hidden Markov Models that integrate TF–target information during model construction, assigns TFs to activation points and models their influence on emission probabilities at later time points, generates combinatorial TF interaction predictions, and was evaluated on multiple mouse and human scRNA-Seq datasets.