HarmonizR

HarmonizR enables batch effect removal in multi-experiment omics datasets containing missing values by adjusting data for compatibility with algorithms such as ComBat and limma.

Key Features:

• Blocking Strategy: A novel blocking strategy reduces runtime while maintaining support for parallel architectures.
• Feature Rescue: A "unique removal" strategy retains additional features for adjustment, achieving feature rescue of up to 103.9% in tested datasets.
• Missing Value Tolerance: Adjusts data to make batch effect removal algorithms functional despite inherent and integration-introduced missing values in (single-cell) RNA, proteomics, and other omics data.

Scientific Applications:

• Multi-experiment Omics Integration: Batch effect reduction for integrated datasets from (single-cell) RNA, proteomics, and other omics data with complex experimental designs.
• Rare Tumor Entity Analysis: Dataset integration specifically developed for rare tumor entities where data sparsity and batch effects complicate statistical analysis.

Methodology:

HarmonizR implements data adjustment with blocking strategies and feature rescue techniques to prepare data for batch effect removal algorithms; the blocking strategy partitions data to improve runtime efficiency, while the "unique removal" strategy preserves features that would otherwise be excluded due to missing value patterns.