AWST

AWST performs asymmetric Winsorization per sample to normalize RNA-seq expression data for variance stabilization and improved unsupervised clustering.

Key Features:

• Robustness to Data Perturbations: AWST reduces the impact of data perturbations by adjusting extreme expression values on a per-sample basis.
• Elimination of Pre-Clustering Gene Selection: AWST obviates the need to pre-select the most informative genes prior to sample clustering.
• Biologically Meaningful Clusters: AWST produces clusters that preserve biological signal in both bulk RNA-seq and single-cell RNA-seq datasets.

Scientific Applications:

• Bulk RNA-seq analysis: Normalization for bulk RNA-seq to support accurate unsupervised clustering.
• Single-cell RNA-seq analysis: Normalization for single-cell RNA-seq to support accurate unsupervised clustering.
• Unsupervised clustering: Preprocessing to improve performance and biological interpretability of unsupervised clustering of samples or cells.

Methodology:

Asymmetric Winsorization applied per sample adjusts extreme values in the data distribution to stabilize variance across samples.