spqn

spqn implements spatial quantile normalization (SpQN) to remove the mean-correlation relationship in correlation matrices built from gene expression data, serving as a pre-processing step for co-expression analysis.

Key Features:

• Mean-correlation bias removal: Spatial quantile normalization removes the dependence between correlation estimates and gene expression levels in RNA-seq data, both bulk and single-cell.
• Local distribution normalization: The method normalizes local distributions within a correlation matrix to correct expression bias in network reconstruction.
• Co-expression preprocessing: SpQN functions as a pre-processing step prior to co-expression analysis to prevent missing biologically relevant lowly expressed gene pairs such as transcription factors.

Scientific Applications:

• Co-expression network analysis: SpQN corrects expression bias in gene network reconstruction to prevent highly expressed genes from being artificially more likely to appear highly correlated.
• Transcription factor identification: The method enables detection of potentially biologically relevant but lowly expressed gene pairs, including transcription factors, that would otherwise be missed due to mean-correlation bias.
• RNA-seq data analysis: SpQN addresses mean-correlation relationships observed in both bulk and single-cell RNA-seq datasets.

Methodology:

Spatial quantile normalization (SpQN) normalizes local distributions in a correlation matrix to remove the mean-correlation relationship.