Benchmark cancer datasets for Clustering algorithms for Omics-based Patient Stratification (COPS)

This repository contains seven multi-omic cancer datasets including several cancer types (breast, kidney, lung, ovary, prostate, and thyroid cancers as well as low grade gliomas) that were used for benchmarking several multi-view clustering algorithms implemented by COPS (https://github.com/UEFBiomedicalInformaticsLab/COPS). The datasets were originally compiled from The Cancer Genoma Atlas (TCGA) and downloaded using the curatedTCGAData R-package. The datasets include copy-number variations, methylomics as well as mRNA and miRNA transcriptomics. The methylomics data was mapped to genes by averaging methylation level of probes associated with the promoter regions of genes. Similarly the miRNA transcriptomics data was mapped to genes by using known and predicted miRNA -> gene interactions. Updated survival data was acquired from the Liu et al. 2018 paper. This repository also includes two sets of cancer associated pathway networks used by pathway-based multi-omic methods benchmarked in our study. NCI-PID pathways were downloaded using the ndexr R-package on December 22 2021. While KEGG pathways were downloaded using the pathview R-package on May 3 2022. More details on the processing can be found on the related publication.