A Note on the Evaluation of Mutation Prioritization Algorithms

Recent developments in the field of gene sequencing technology greatly accelerated discovery of mutations that cause various genetic disorders. At the same time, a typical sequencing experiment generates a large number of candidate mutations, hence detecting single or few causative variants is still a formidable problem. Many computational methods have been proposed to assist this process, from which a large portion employ statistical learning in some form. Consequently, each newly designed algorithm is routinely compared to other competing systems in hope to demonstrate advantageous performance. In this work we review and discuss several issues related to the current practice of evaluation of mutation prioritization algorithms and suggest possible directions for improvements.