Applied Math Colloquium Talk - Dr. Art Poon

Speaker: Art Poon

Title: Outbreak detection with Markov-modulated Poisson processes

Abstract:
Genetic distance functions measure the difference between two sequences that have accumulated changes since their descent from a common ancestor.  Hierarchical clustering of these distances provide an efficient method for reconstructing evolutionary histories --- how species or populations are related by common ancestors.  For infectious diseases, these distances are also used to detect outbreaks.  Sampling infections that retain a high degree of genetic similarity implies that they are related by recent transmission events.  Nonparametric clustering on large sequence databases has become an enormously popular application of genetic distances to detect potential outbreaks, because they can potentially be used to guide outbreak management efforts in near real time.  However, this application is fraught with severe biases that the field has only begun to identify.  We propose a model-based clustering approach using Markov-modulated Poisson processes to detect transmission outbreaks and use simulations to compare this approach to current nonparametric methods.