Profile
Project Title: Deep learning algorithms to detect signals of adaptation from genomic data
Summary: Many phenotypic variations between populations and species are a result of alleles being maintained, lost, or fixed through generational time points. Simulations of demographic history now play an important role in helping assess evolutionary models, as well as providing a rich treasure trove of data. With population genetics transitioning into a data-driven discipline, new and emerging methods, particularly in the field of machine learning, are being developed for the analysis of genetic inference. With previous methods of detecting selection through likelihood and Bayesian approaches slowly becoming unfeasible, methods of machine learning involve utilising synthetic data in order to train neural networks to detect areas of the genome displaying patterns of selection.
A framework will be developed that provides evolutionary simulated data to train a deep learning model in order to detect windows of the genome that display selective sweeps. Through this, signatures of selection can be detected and investigated further through developmental methods for the study of evolutionary history and future.
Supervisor: