Thesis Presentation: Jake Roberts
Title: Likelihood-free Inference for Systemic AA Amyloidosis Dynamics in Captive Cheetahs
Abstract: Captive cheetah populations exhibit a high prevalence of systemic AA amyloidosis, a prion-like disease characterized by the deposition of serum amyloid A protein into tissue matter. The disease is considered a leading cause of mortality and morbidity among captive cheetahs, but little is known about the exact cause. Previous research hypothesizes genetic transmission components, transmission through direct contact with amyloid fibrils, inflammatory catalysts, or a combination of these factors. Three agent-based models were developed reflecting each hypothesized mechanism. The models were fitted to data from pathology reports and cheetah ancestry. Due to the complex structure and interaction between the latent disease states in the population, the model likelihoods were unavailable, and approximate Bayesian computation was used to obtain approximate posterior samples. Models were compared in terms of out-of-sample prediction accuracy. We find evidence of external, non-genetic mechanisms for AA amyloidosis in captive cheetahs, but pinpointing this mechanism requires additional targeted data collection.
Advisor: Oksana Chkrebtii
