Using iterative cluster merging with improved gap statistics to perform online phenotype discovery in the context of high-throug