Exploratory analysis of multiple omics datasets using the adjusted RV coefficient.

Area of Science:

• Bioinformatics
• Systems Biology
• Multivariate Statistics

Background:

• Integrating multiple high-dimensional omics datasets is crucial for systems biology but challenging.
• Existing multivariate statistical tools often handle only two datasets, limiting analysis of larger experimental outputs.
• High dimensionality in omics data introduces bias to standard RV (Robust Variance) coefficient calculations.

Purpose of the Study:

• To develop a robust statistical method for analyzing similarity among multiple high-dimensional omics datasets.
• To address the bias issue in the RV coefficient when applied to high-dimensional data.
• To provide a reliable tool for explorative analysis of omics data integration in systems biology.

Main Methods:

• Introduced an adjusted RV coefficient, an unbiased version of the RV coefficient for independent datasets.
• Evaluated the adjusted RV's performance against existing RV versions using mean square error and independence test power.
• Applied the adjusted RV to a dataset of 19 multivariate datasets from a systems biology experiment.

Main Results:

• The adjusted RV demonstrates improved performance as an estimator, especially for very high-dimensional data.
• Pairwise adjusted RV values generated a similarity matrix effectively.
• This similarity matrix, when used with hierarchical clustering or multidimensional scaling, revealed biologically meaningful subgroups of datasets.

Conclusions:

• The adjusted RV is a valuable and unbiased statistical tool for assessing similarity between multiple high-dimensional omics datasets.
• This method facilitates effective data integration and pattern discovery in complex systems biology research.
• The adjusted RV enables the identification of biologically relevant data structures within large-scale experiments.