distnet

distnet evaluates the fidelity of low-dimensional visualizations by comparing two-dimensional placements to original high-dimensional similarities as an R package for validating dimension-reduction results.

Key Features:

• Interactive visualization: Provides interactive inspection of embeddings to link two-dimensional placements with original high-dimensional relationships.
• Assessment of dimension-reducing plots: Identifies discrepancies between two-dimensional placements and similarities in the original feature space to detect visualization artifacts.
• Complementary tool integration: Operates alongside focusedMDS to provide a broader assessment of visualizations and to support analyses relevant to personalized medicine.

Scientific Applications:

• High-dimensional data analysis: Applies to high-throughput assay datasets where dimension reduction is required to interpret complex measurements.
• Validation of biological insights: Confirms that observed clusters or patterns in reduced dimensions reflect true relationships rather than projection artifacts.
• Personalized medicine: Supports individualized data interpretation when used with focusedMDS to examine specific data-point relationships within broader cohorts.

Methodology:

Performs a direct comparison between the spatial arrangement of points in reduced-dimension embeddings and pairwise similarities in the original high-dimensional feature space.