A new method for combining different types of biological data
The baRRR method integrates data across multiple sample groups and different sources to find patterns.
Researchers introduced a method called Bidimensional Augmented Reduced Rank Regression (baRRR). This tool integrates data across multiple cohorts and views.
A cohort is a sample group. A view is a source or modality of data. The method models effects that are related to covariates and effects that are unrelated.
The approach decomposes effect levels into global, partially shared, and individual levels. It works across both dimensions of the data.
The method was tested using data from The Cancer Genome Atlas (TCGA). It used somatic mutations as covariates to find patterns across cancer types and omics platforms.
Simulations show the method has advantages in signal recovery, outcome prediction, and missing data imputation. It is a robust and adaptable tool for integrative analysis.
Key facts
- The baRRR method models covariate-related and unrelated effects at the same time.
- The method decomposes effects into global, partially shared, and individual levels.
- The approach was applied to TCGA data using somatic mutations as covariates.