ASTER

ASTER estimates the number of cell types present in single-cell chromatin accessibility sequencing (scCAS) data to quantify cellular heterogeneity in epigenomic studies.

Key Features:

• Ensemble Learning Approach: Employs an ensemble learning framework that aggregates predictions from multiple models to improve accuracy and reduce bias and variance compared to traditional baseline methods.
• Systematic Evaluation: Evaluated on 27 distinct datasets spanning diverse protocols, sizes, numbers of cell types, degrees of cell-type imbalance, and variations in cell states and qualities.
• Robustness to Technical Variation: Designed to handle variability in scCAS data including differences in sequencing depth, cell-type proportions, and technical noise.

Scientific Applications:

• Cell type number estimation: Provides quantitative estimates of cell type counts in scCAS datasets to support characterization of cellular composition.
• Novel cell type discovery: Supports downstream identification of previously uncharacterized cell types by informing clustering and annotation workflows.
• Cellular differentiation analysis: Aids studies of differentiation and cell-state transitions by supplying estimates of cell-type diversity and composition.
• Regulatory landscape exploration: Facilitates interpretation of chromatin accessibility data to examine gene regulation mechanisms across cell types and conditions.

Methodology:

Uses an ensemble learning framework that aggregates predictions from multiple models to estimate cell-type numbers and address scCAS data variability, including sequencing depth, cell-type proportions, and technical noise.