svapls

svapls adjusts for hidden sources of variability in gene expression studies by extracting sample-specific heterogeneity signals to improve the accuracy of differential expression testing.

Key Features:

• Identification of Hidden Sources of Heterogeneity: Identifies unknown sample-specific sources of heterogeneity, including subject-specific effects and residual variability, that can confound gene expression measurements.
• Partial Least Squares Regression: Uses Partial Least Squares (PLS) regression to extract hidden signals associated with sample-specific heterogeneity.
• Improved Detection Power: Demonstrates improved detection power and reduced error rates through sensitivity analyses on simulated data with diverse patterns of hidden variation.
• Application to Real Datasets and Batch Effect Detection: Validates performance on real-life datasets such as the Golub dataset and identifies potential batch effects in gene expression data.

Scientific Applications:

• Differential Gene Expression Testing: Improves the accuracy of differential gene expression testing by adjusting for hidden variability.
• Genome-wide Testing and Transcript Discovery: Supports genome-wide testing to identify transcripts associated with phenotypes by reducing confounding from sample-specific heterogeneity.
• Batch Effect Detection: Detects potential batch effects in gene expression datasets to inform downstream analyses.

Methodology:

Identifies unknown sample-specific heterogeneity, extracts hidden signals using Partial Least Squares (PLS) regression, performs sensitivity analyses on simulated datasets, and validates findings on the Golub dataset.