jackstraw

jackstraw identifies genomic variables that are significantly associated with principal components (PCs) to determine drivers of systematic variation in large-scale genomic datasets.

Key Features:

• PCA-based analysis: Uses principal component analysis (PCA) to capture latent variation and evaluate associations between genomic variables and PCs.
• Significance testing for PCs: Identifies genomic variables significantly associated with any subset or linear combination of PCs.
• Over-fitting mitigation: Employs a statistical framework that mitigates over-fitting arising from deriving PCs from the same genomic variables.
• Accurate significance measures: Produces calibrated measures of statistical significance as demonstrated by simulation studies.
• Large-scale applicability: Applies to large-scale genomic datasets to link variable-level signals to systematic variation.
• Simplified hypothesis testing: Addresses complex significance testing problems for associations with latent factors.

Scientific Applications:

• Yeast cell-cycle gene detection: Identified cell-cycle regulated genes in yeast cell-cycle gene expression data.
• Post-trauma gene expression analysis: Applied to post-trauma patient gene expression data to reveal greater enrichment for inflammatory-related gene sets compared to analyses based on clinically defined phenotypes.
• Latent factor association studies: Used to associate genomic variables with latent factors in complex biological datasets.

Methodology:

Applies principal component analysis (PCA) to capture systematic variation and uses a statistical framework that mitigates over-fitting to test significance of associations between genomic variables and PCs, with validation via extensive simulation studies.