AC-PCoA

AC-PCoA adjusts for confounding factors in biological datasets by applying Principal Coordinate Analysis (PCoA) to distance matrices to derive low-dimensional representations that minimize associations with technical, population-structure, or experimental confounders.

Key Features:

• Implementation: Provided as an R package for computational analysis of biological data.
• Dimensionality Reduction: Employs Principal Coordinate Analysis (PCoA) to reduce dataset dimensionality while preserving structure from multiple distance measures.
• Confounding Factor Adjustment: Adjusts lower-dimensional representations to minimize associations with confounders such as technical variation, population structure, or experimental conditions.
• Multiple Distance Measures: Operates on various distance metrics beyond Euclidean, enabling analysis of diverse data types including sequencing-derived distances.
• Visualization and Analysis: Produces adjusted low-dimensional embeddings that improve clarity and interpretation of visualizations.
• Statistical Testing and Clustering: Enhances the reliability of statistical tests and clustering results by reducing confounding bias.
• Classification and Prediction: Supports downstream classification and prediction tasks using adjusted representations.

Scientific Applications:

• Genomics: Applied to genomic datasets to remove confounding influences and reveal true biological variation.
• Transcriptomics: Applied to transcriptomic data to adjust for technical and experimental confounders.
• Other omics sciences: Applicable across other omics disciplines where distance-based analyses are used.
• Sequencing data: Handles complex sequencing-derived distances through flexible distance metrics and PCoA.
• Downstream analyses: Improves visualization, statistical testing, clustering, classification, and prediction on simulated and real datasets by mitigating confounding effects.

Methodology:

Applies Principal Coordinate Analysis to distance matrices and adjusts the resulting low-dimensional representations to minimize associations with specified confounding factors.