BCseq (bias-corrected sequencing analysis)

BCseq performs bias-corrected quantification of gene expression from single-cell RNA sequencing (scRNA-seq) data to reduce technical noise and recover dropout events for more accurate single-cell expression estimates.

Key Features:

• Bias Correction: Implements a data-adaptive model to correct inherent biases in scRNA-seq measurements across datasets.
• Technical Noise Reduction: Applies advanced noise-reduction techniques to decrease technical variability that obscures biological signal.
• Dropout Rescue: Infers missing gene expression values by leveraging information from similar cells with a weighted consideration approach.
• Nonlinear Sequencing-Depth Weighting: Uses a nonlinear weighting scheme so cells with higher sequencing depth contribute more to quantification.
• Quality Scoring: Assigns a per-gene per-cell quality score to quantify confidence in each expression measurement.

Scientific Applications:

• Detection of Cell Subtypes: Improves identification of distinct cell subtypes by providing more accurate single-cell gene expression estimates.
• Tracing Cell Transitions: Supports analyses of cellular transitions during development and disease by reliably quantifying expression changes.
• Differential Gene Expression Analysis: Increases robustness in detecting differentially expressed genes from scRNA-seq datasets.
• Cell Subtype Classification: Enhances classification of cellular heterogeneity through higher-confidence expression measurements.

Methodology:

BCseq integrates data-adaptive bias correction and advanced noise-reduction techniques, leverages information from similar cells to infer dropouts, applies a nonlinear sequencing-depth weighting scheme, and assigns per-gene per-cell quality scores.