Investigating the Contribution of Morphological Divergence to Behaviour using AI

Identifying the genes underlying phenotypic divergence is crucial to understanding how new phenotypes arise and species differentiate. Recently, we identified some of the genes that contribute to divergence in size and shape of male genitalia between two species of fruit flies (Hagen et al., 2019, 2021). Genital structures evolve very quickly between closely related species (Eberhard, 2010) and are known to affect mating behaviour and reproductive fitness (Frazee and Masly, 2015, LeVasseur-Viens et al., 2015).

However, identifying behavioural differences encoded by the effect of these genes on male genital morphology necessitates large-scale experiments under tightly controlled conditions. Recent advances in automated tracking and behaviour annotation make high throughput analysis of behaviour now possible.

Director of Studies: Dr Daniela Santos-Nunes 

Other Supervisors: Prof Fabio Cuzzolin

This project aims to implement a high-throughput pipeline for mating behaviour analysis, using machine-learning techniques such as deep neural networks, to analyse video data acquired from custom-made behaviour arenas.

Objective 1: To use males that differ in the species-specific allele they carry for each of the focal genes to determine if their effect on genital morphology results in differences in stereotypical mating, including the nature and timing of those differences.

Objective 2: To automate the discovery of new behavioural phenotypes and apply the pipeline in more complex contexts (e.g. multiple mating pairs), leading to better understanding of group mating behaviour in the wild. Results will provide fundamental insight into the proximate causes of behavioural evolution driving speciation.

There is an additional requirement to undertake up to 6 hours undergraduate teaching/week during semesters and to participate in a teaching skills course without further remuneration.