Machine learning reveals singing rhythms of male Pacific field crickets are clock controlled.

Abstract

Circadian rhythms are ubiquitous in nature and endogenous circadian clocks drive the daily expression of many fitness-related behaviors. However, little is known about whether such traits are targets of selection imposed by natural enemies. In Hawaiian populations of the nocturnally active Pacific field cricket (Teleogryllus oceanicus), males sing to attract mates, yet sexually selected singing rhythms are also subject to natural selection from the acoustically orienting and deadly parasitoid fly, Ormia ochracea. Here, we use T. oceanicus to test whether singing rhythms are endogenous and scheduled by circadian clocks, making them possible targets of selection imposed by flies. We also develop a novel audio-to-circadian analysis pipeline, capable of extracting useful parameters from which to train machine learning algorithms and process large quantities of audio data. Singing rhythms fulfilled all criteria for endogenous circadian clock control, including being driven by photoschedule, self-sustained periodicity of approximately 24 h, and being robust to variation in temperature. Furthermore, singing rhythms varied across individuals, which might suggest genetic variation on which natural and sexual selection pressures can act. Sexual signals and ornaments are well-known targets of selection by natural enemies, but our findings indicate that the circadian timing of those traits' expression may also determine fitness.