awst

awst regularizes RNA-seq read counts through an Asymmetric Within-Sample Transformation to reduce noise effects on sample classification.

Key Features:

• Standardization: Per-sample standardization of gene expression data as the first main step of the transformation.
• Smoothing: Asymmetric winsorization that reduces noise in lowly expressed features affected by background effects and low signal-to-noise ratio, and diminishes influence of highly expressed features resulting from amplification bias and experimental artifacts.
• Robustness to perturbations: The transformation is robust to data perturbations and eliminates the need for selecting the most informative genes prior to sample clustering.
• Applicability to bulk and single-cell RNA-seq: Produces robust and biologically meaningful clusters in both bulk and single-cell RNA-seq applications.

Scientific Applications:

• RNA-seq sample clustering: Accurate unsupervised clustering of RNA-seq expression profiles without requiring prior gene selection.
• Noise reduction in gene expression data: Regularization of RNA-seq read counts to mitigate background effects, low signal-to-noise ratio, and amplification bias.
• Single-cell RNA-seq analysis: Clustering of single-cell RNA-seq expression profiles into biologically meaningful groups.

Methodology:

AWST comprises two main steps: standardization and smoothing. The procedure applies per-sample standardization followed by asymmetric winsorization to transform gene expression data.