Benchmark study of feature selection strategies for multi-omics data

These data sets are the pre-processed versions of the multi-omics data sets used in the benchmark study presented in the paper "Benchmark study of feature selection strategies for multi-omics data" by Yingxia Li, Ulrich Mansmann, Shangming Du, and Roman Hornung. The outcome feature is "TP53_mutation" in each data set, where "1" / "0" indicates the presence / absence of a TP53 mutation in the respective patients. The remaining features are clinical and omics features, where the suffix "_clinical" indicates clinical features, the suffix "_cnv" copy number variation features, the suffix "_mirna" miRNA features, the suffix "_mutation" mutation features, and the suffix " _rna" RNA features. Note that while predicting the outcome feature TP53 yes vs. no is not meaningful contextually, TP53 mutations have been found to be associated with poor clinical outcomes in cancer patients [1]. Against this background, TP53 can be used as a surrogate for a phenotypic outcome. Thus, these data sets are meant for testing machine learning or statistical procedures, they may not be useful for biological analysis.