Evolutionary prediction for new echolocators

It was suggested over 60 years ago that echolocation is widespread in animals. Although echolocation has been found to evolve independently in several mammalian lineages, this hypothesis remains largely untested due to the difficulty of determining the presence of echolocation. To address this issue, we combined molecular convergence analyses from 190 high-quality mammalian genomes with machine learning to predict potential new mammalian echolocators. Our model predicted three promising lineages of echolocating mammals. Behavioral experiments confirmed that the gracile shrew mole (Uropsilus gracilis), the highest- ranking species among predicted echolocators, as well as almost all shrew moles (Uropsilinae), are capable of echolocation through the use of ultrasonic pulses. In contrast to most bats that use laryngeal echolocation, the characteristics of calls, the morphology of the stylohyal bone, and the results of tongue ligation experiments all suggest that shrew moles produce ultrasonic clicks with their tongues for echolocation. Finally, we estimated at least 20% of all living mammalian orders with echolocation ability, thus empirically supporting Griffin’s hypothesis that echolocation is widespread among animals. Our findings not only provide evidence that three novel lineages of echolocating mammals, but also demonstrate that phylogenetically replicated phenotypes can be predicted through genetic convergence.