cisTopic

cisTopic applies probabilistic topic modeling to single-cell ATAC-seq data to identify coaccessible enhancers and stable cell states in sparse epigenomic datasets.

Key Features:

• Simultaneous discovery: Identifies coaccessible enhancers and stable cell states concurrently from single-cell ATAC-seq data.
• Probabilistic topic modeling: Uses a probabilistic topic modeling framework to extract regulatory topics from sparse single-cell epigenomic datasets.
• Cell type, enhancer and TF identification: Facilitates robust identification of distinct cell types, enhancer regions, and associated transcription factors.
• Application to diverse datasets: Demonstrated on datasets including differentiating hematopoietic cells, brain tissues, and transcription factor perturbations.
• Regulatory heterogeneity insight: Models enhancer coaccessibility and their association with specific cell states to reveal regulatory heterogeneity.
• Accelerated inference: cisTopic v3 incorporates WarpLDA for faster topic modeling inference.

Scientific Applications:

• Cell type identification: Classifies cell types based on single-cell ATAC-seq epigenomic signatures.
• Enhancer mapping: Maps enhancers that are coaccessible within specific cell states.
• Transcription factor analysis: Identifies transcription factors associated with particular enhancers and cell states.
• Study of regulatory mechanisms: Analyzes regulatory heterogeneity and mechanisms underlying cellular differentiation and responses to transcription factor perturbations.

Methodology:

cisTopic applies a probabilistic topic modeling framework to single-cell ATAC-seq data, extracting topics from sparse matrices and using WarpLDA in cisTopic v3 for faster topic inference.