Benchmark study of feature selection strategies for multi-omics data