Abstract
Genetic subtyping of viruses and bacteria is a critical tool for visualizing and modeling their geographic distribution and temporal dynamics. Quantifying viral dynamics is of particular importance for the novel coronavirus responsible for COVID-19, SARS-CoV-2. Effective containment strategies and potential future therapeutic and vaccine strategies will likely require a precise and quantitative understanding of viral transmission and evolution. In this paper, we employ an entropy-based analysis to identify mutational signatures of SARS-CoV-2 strains in the GISAID database available as of April 5, 2020. Our analysis method identifies nucleotide sites within the viral genome which are highly informative of variation between the viral genomes sequenced in different individuals. These sites are used to characterize individual virus sequence with a characteristic Informative Subtype Marker (ISM). The ISMs provide signatures that can be efficiently and rapidly utilized to quantitatively trace viral dynamics through geography and time. We show that by analyzing the ISM of currently available SARS-CoV-2 sequences, we are able to profile international and interregional differences in viral subtype, and visualize the emergence of viral subtypes in different countries over time. To validate and demonstrate the utility of ISM-based subtyping: (1) We show the distinct genetic subtypes of European infections, in which early on infections are related to the viral subtypes that has become dominant in Italy followed by the development of local subtypes, (2) We distinguish subtypes associated with outbreaks in distinct parts of the United States, identify the development of a local subtype potentially due to community to transmission and distinguish it from the predominant subtype in New York, suggesting that the outbreak in New York is linked to imported cases from Europe. (3) We present results that quantitatively show the temporal behavior of the emergence of SARS-CoV-2 from localization in China to a pattern of distinct regional subtypes as the virus spreads throughout the world over time. Accordingly, we show that genetic subtyping using entropy-based ISMs can play an important complementary role to phylogenetic tree-based analysis, such as the Nextstrain project, in efficiently quantifying SARS-CoV-2 dynamics to enable modeling, data-mining, and machine learning tools. Following from this initial study, we have developed a pipeline to dynamically generate ISMs for newly added SARS-CoV-2 sequences and generate updated visualization of geographical and temporal dynamics, and made it available on Github at https://github.com/EESI/ISM.