Benchmarking relief-based feature selection methods for bioinformatics data mining.

•Relief-based feature selection (RBAs) efficiently detect feature interactions.•RBAs handle genetic heterogeneity, missing/imbalanced data, and quantitative traits.•SURF∗ and MultiSURF∗ are not suited to detecting main effects.•The new MultiSURF algorithm performs most consistently over different problems.•ReBATE software offers easy access to multiple, flexible RBAs.