A1303
Title: PREMISE: A quality-aware probabilistic framework for high resolution pathogen resolution in diagnostic sequence samples
Authors: Karin Dorman - Iowa State University (United States) [presenting]
Sriram Vijendran - Iowa State University (United States)
Tavis Anderson - National Animal Disease Center - Agricultural Research Service - United States Department of Agriculture (United States)
Oliver Eulenstein - Iowa State University (United States)
Abstract: The rapid and accurate identification of viral pathogens In diagnostic and surveillance samples is critical for public health risk assessment and intervention. Some high-threat virus species are now so thoroughly monitored by genome Sequencing that nearly all circulating genotypes are recorded In reference databases. Current viral identification methods presume noisy sequence data and incomplete reference databases to make low-Resolution identification, prioritizing scalability by discarding variant linkage and read quality information as mostly uninformative. Yet In well-characterized viral systems, it may now be feasible to substantially boost identification accuracy while simultaneously offering uncertainty quantification. Thus, PREMISE (Pathogen Resolution via Expectation Maximization In Sequencing Experiments) is proposed. PREMISE combines advanced string data structures for efficient alignment with a quality-aware probability model for high-Resolution viral genome identification. Validated against simulated and empirical datasets, PREMISE more accurately identifies source strains and their abundance while achieving more precise read assignments than competing methods. Ultimately, this framework represents a significant advance In viral identification, providing a foundation for extensions to better error models, methods to detect novel variants, and statistical corrections for sampling bias In diagnostics.
