Machine learning and genomics: Using neural networks for population assignment of a threatened seabird Population genomics is aiding researchers in uncovering more information about non-model organisms, and has aided in the development of species and population-specific conservation plans. For many species, effective conservation remains difficult due to low population genetic structure and difficulty in accurately assessing potential threats. The Leach’s storm-petrel (Hydrobates leucorhous) is a migratory pelagic seabird that breeds in large colonies throughout the North Pacific and Atlantic Oceans. In the past 50 years, Atlantic populations have declined by an estimated 54%. Several potential threats, ranging from offshore structures to increased predation from gulls, have been identified, however determining the impact of these widely distributed threats on specific colonies remains difficult due to the species low genetic structure and migratory behaviour. Where previous population assignment studies on this species have failed, I aim to use a combination of genomic data and novel machine learning methods to investigate the genetic structure of Atlantic Leach’s Storm-Petrels and assign individuals of unknown origin to their respective breeding colonies. Using DNA collected from over 300 individuals from 11 different populations, as well as 84 deceased individuals of unknown origin, I attempted to use the novel neural network popfinder to generate colony of origin predictions and determine what threats appear to be the most pertinent to specific colonies.
Comments are closed.
|
Categories
All
Archives
April 2024
|