integration_analysis_scripts

Scripts for multi-omics integration

Unsupervised analysis: integration_unsupervised.R

This script performs unsupervised analyses (clustering) from transformed expression data (e.g., log fpkm) and methylation beta values

Prerequisites

This R script requires the following packages:

• iClusterPlus
• gplots
• lattice

Usage

Rscript integration_unsupervised.R [options]

PARAMETER	DEFAULT	DESCRIPTION
-d	NULL	File with somatic mutation data
-C	NULL	File with copy number variation data
-r	NULL	File with expression data
-m	NULL	File with methylation data (beta values)
-k	2	Minimum number of clusters
-K	6	Maximum number of clusters
-c	2	Number of cores
-o	out	output prefix
-h		Show help message and exit

For example, one can type

Rscript integration_unsupervised.R -r expression_matrix.txt -o output/

Details

The script involves 3 steps

• Data transformation of methylation beta values, using the logit function
• Clustering across a range of LASSO lambda penalties and for each number of clusters K using iClusterPlus
• Selection of the best lambda value (BIC) for each K, and plot of the R^2 as a function of K to help the choice of K
• Selection of the top features differentiating the clusters

Output

• A figure with R^2 as a function of K, and cluster memberships of each sample as a function of K

In addition, for each value of K:

• an .RData file with clustering results
• a heatmap with the top features for each dataset
• a .txt file with the name of the top features for each dataset

Regression analysis for unsupervised analysis: PCA_regression.R

This script provides functions to perform regression analysis between variables (e.g., batch variables or clinical variables) and latent factors as obtained by PCA or group factor analysis.