pcqc

pcqc selects principal components that capture informative variance in single-cell RNA sequencing (scRNA-seq) data to improve clustering and the detection of rare cell populations.

Key Features:

• Optimal Principal Component Selection: Uses a quantile-based criterion focusing on the tails of the per-observation distribution of variance explained to prioritize components that capture signals from rare cell populations.
• Computational Efficiency: Offers a computationally efficient alternative to existing selection methods for large-scale scRNA-seq datasets.
• Robustness Across Datasets: Evaluated across three single-cell RNA-sequencing datasets and shown to match or outperform traditional variance-explained selection criteria.

Scientific Applications:

• Preprocessing for scRNA-seq: Selects principal components for downstream clustering and analysis in single-cell RNA-seq workflows.
• Dimensionality Reduction and Visualization: Improves inputs for non-linear dimensionality reduction methods such as t-SNE and UMAP to yield more accurate representations of cell populations.
• Rare Cell Population Detection: Enhances identification of rare cell subtypes that may be underrepresented by traditional variance-explained approaches.

Methodology:

pcqc applies a quantile-based approach to the distribution of variance explained per observation to identify and select principal components emphasizing signal in the distribution tails.