GSEAlm

GSEAlm integrates linear modeling diagnostics with gene-set enrichment analysis to improve detection and interpretation of gene expression patterns in transcriptomic datasets.

Key Features:

• Linear model fitting: Fits linear models to gene expression data to quantify associations between covariates and gene-level expression.
• Regression diagnostics: Computes and utilizes various regression diagnostics to assess influence and residual behavior for individual genes and samples.
• Outlier and influential sample detection: Identifies outliers and influential samples via diagnostic measures to highlight problematic observations.
• Model fit evaluation: Provides assessments of model fit to determine how well linear models represent the underlying biological data.
• Model expansion exploration: Offers diagnostic insights that suggest potential areas for model expansion or refinement.
• Chromosome-band GSEA support: Supports gene-set enrichment analysis based on chromosome-band mapping of genes.
• Residual analysis by loci: Enables analysis of individual residuals grouped by chromosomal loci to detect locus-specific patterns.

Scientific Applications:

• ALL dataset analysis: Demonstrated on an adult acute lymphoblastic leukemia (ALL) dataset for real-world evaluation of methods.
• Problematic sample identification: Uses residuals grouped by chromosomal loci to identify problematic samples and potential data-entry errors.
• Hyperdiploidy detection: Identified hyperdiploidy as a significant factor influencing gene expression and as an indicator of suspected DNA copy number abnormalities.
• Chromosomal expression analysis: Revealed significant expression differences between hyperdiploid and diploid groups across chromosomes X, 21, 14, 19, 22, 3, and 13, including differences not associated with copy number changes.

Methodology:

Fitting linear models to gene expression data and computing regression diagnostics, including analysis of individual residuals grouped by chromosomal loci and GSEA based on chromosome-band mapping.