Quantifying Pathogen Surveillance Using Temporal Genomic Data

Chan, Joseph M.; Rabadan, Raul

In this paper, we propose a readily interpretable and computable quantitative measurement for genomic surveillance of a pathogen that directly accounts for the number of isolates, the evolutionary rate, and the time of sample collection without the need to define arbitrary clades or species or the need for a full phylogenetic reconstruction. This measure ranks a surveillance system as more complete if it is able to capture a greater proportion of the pathogen's diversity. We apply this measure to influenza virus and compare the surveillance of different influenza virus strains in different hosts and geographic regions. We find that, compared to human seasonal strains, sampling is indeed substantially lower for swine H1N1 and avian non-H5N1 influenza virus, historically overlooked strains despite their pandemic potential. We also find that avian H5N1 influenza virus surveillance in the WHO transmission zones of northern and western Africa; eastern, southern, and southeastern Asia; and eastern, southwestern, and northern Europe is high and may potentially serve as an effective early warning system given a list of genetic determinants of mammalian adaptation. Avian H5N1 influenza virus surveillance in North America, however, is much less comprehensive. We similarly apply our methodology to other RNA viruses and show inadequate surveillance of both dengue and West Nile viruses. Finally, we perform a comparative analysis of the q2 coefficient and other methods, particularly phylogenetic and clustering alternatives, and find that the q2 coefficient produces similar results with negligible computation time.