Comments on "leave-cluster-out cross-validation is appropriate for scoring functions derived from diverse protein data sets": significance for the validation of scoring functions.