UNMF

UNMF performs unified nonnegative matrix factorization (NMF) to extract informative nonnegative components from multi-dimensional omics datasets represented as tidy data or tensors and is implemented in the R package moltenNMF.

Key Features:

• Tidy Data Compatibility: Accepts tidy data formats to apply NMF directly without mandatory reshaping into alternative matrix forms.
• Missing Values and Repeated Measurements: Handles datasets containing missing values and repeated measurements common in biological experiments.
• Tensor Integration: Operates on tensor-form data to model and decompose multi-dimensional omics structures.
• Unified Analytical Framework: Provides a unified approach to nonnegative matrix factorization across different data formats and structures.
• R Implementation (moltenNMF): Implemented in the R package moltenNMF.

Scientific Applications:

• Pattern Extraction: Extracts informative patterns and components from complex biological and omics datasets.
• Biomarker Discovery: Identifies candidate biomarkers from multi-dimensional omics data.
• Multi-omics Integration: Integrates diverse omics data types represented as tensors or tidy tables.
• Mechanism Inference: Supports inference of underlying biological mechanisms via component analysis.

Methodology:

Extends nonnegative matrix factorization (NMF) to decompose high-dimensional data into nonnegative components while accommodating tidy data formats, tensor structures, missing values, and repeated measurements.