Principal Component Variation Association
pc_var_association.RdQuantifies the association between sample-level metadata covariates and factors and principal components (PCs) derived from a bulk RNA-seq expression dataset. The function evaluates both categorical variables (covariates) and continuous variables across the top principal components and returns a matrix of statistical significance values.
This function is a custom implementation inspired by Principal Variance Component Analysis (PVCA) described by Li, J., Bushel, P., et. al.Chapter 12, 2010. It can be used both before and after batch correction to evaluate whether experimental, biological, or technical variables explain major sources of variation in the dataset.
Value
A matrix containing p-values representing the association between each metadata variable and each principal component. The matrix has principal components as rows and metadata variables as columns.
A heatmap visualizing the resulting p-value matrix is also generated.
Details
INPUT:
A DESeqDataSet object containing normalized or appropriately
transformed expression data and corresponding sample metadata.
Metadata variables are evaluated according to their data type:
Categorical variables are evaluated using ANOVA and an F-test.
Continuous variables are evaluated using an appropriate linear model or association test.
References
Li, J., Bushel, P., Chu, T., Wolfinger, R., Batch Effects and Noise in Microarray Experiment, Chapter 12, 2010