Jinyuan Guo
BME PhD Proposal Presentation

Date: 2025-03-30
Time: 8:30 (ATL, time) 20:30(Beijing Time)
Location / Meeting Link: https://gatech.zoom.us/j/92268521290?pwd=xZ4Eaf4OmlIZ2aJaI29avPbQCgXWls.1

Committee Members:
Peng Xi; Huaiqiu Zhu; Peng Qiu; Xiaodong Su; Ziding Zhang


Title: Developing Foundation Models for Predicting Viral Animal Host Range in Surveillance of Zoonotic Viruses

Abstract:
Emerging human infectious viruses originating from animals have consistently posed severe threats to global public health, as exemplified by the COVID-19 pandemic and numerous other zoonotic outbreaks. A fundamental barrier to pandemic prevention is our limited understanding of the host range of animal viruses — the diversity of animal species they can infect — which directly shapes the risk of viral spillover into human populations. While previous models have provided valuable insights into viral host prediction, they often suffer from limited host coverage, poor generalization to novel viruses, and a heavy reliance on incomplete phenotypic data. Recent advances in genomic sequencing and foundation models offer new opportunities to predict viral host range directly from viral genetic information. In this dissertation, we aim to systematically predict the animal host ranges by developing advanced foundation models. Specifically, this work is organized into three major aims: (Aim 1) We curate a comprehensive and diverse dataset of virus-animal associations, including mammals, birds, and arthropods, by integrating and harmonizing existing databases; (Aim 2) We developed a foundation model trained on viral genome and protein sequences, and build an ensemble model incorporating viral compositional traits, phylogeny, and protein-protein interaction signals to robustly predict host range;(Aim 3) We point out the limitations of current foundation models by integrating biological insights. Through this research, we expect to achieve accurate, generalizable, and interpretable predictions of viral host ranges, with particular emphasis on novel and understudied viruses. Ultimately, this work aims to provide a practical tool for transforming viral genomic data into actionable insights for outbreak preparedness and early-warning surveillance of zoonotic threats.