nipalsMCIA: flexible multi-block dimensionality reduction in R via nonlinear iterative partial least squares

Area of Science:

• Bioinformatics
• Computational Biology
• Data Science

Background:

• Multi-omics data analysis is crucial for biological insights.
• Unsupervised learning methods are needed for clustering, visualization, and feature selection.
• Existing methods may struggle with large-scale omics datasets.

Purpose of the Study:

• To introduce nipalsMCIA, an efficient implementation of multiple co-inertia analysis (MCIA).
• To provide a robust tool for joint dimensionality reduction in multi-omics data.
• To improve the speed and scalability of unsupervised learning for omics analyses.

Main Methods:

• Implemented Multiple Co-Inertia Analysis (MCIA) using Nonlinear Iterative Partial Least Squares (NIPALS).
• Developed nipalsMCIA as a Bioconductor package.
• Applied the method to both bulk and single-cell omics datasets.

Main Results:

• nipalsMCIA offers significant speed-up for large sample size and/or feature dimension datasets.
• Demonstrated effectiveness in joint dimensionality reduction for multi-omics data.
• Validated performance on both bulk and single-cell omics data.

Conclusions:

• nipalsMCIA provides a scalable and efficient solution for multi-omics data analysis.
• The tool facilitates unsupervised learning tasks including clustering, visualization, and feature selection.
• Available as a Bioconductor package with documentation and vignettes for ease of use.