Speaker
Description
Astroviruses affect human public health and present multiple complexities that challenge taxonomic assignment. Human astrovirus infections (HAstV subgenera) present with gastroenteritis and diarrhea with recent reports of expanded pathologies including neurotropic effects, contributing to encephalitis and meningitis in immunocompromised patients, and mild respiratory effects such as cough and rhinorrhea. Liver and kidney-specific effects have also been observed in other mammals infected with Astroviruses (Mamastroviruses; MAstV genera). Astroviruses have cross-species transmission with infection of other mammals and birds (Avastroviruses; AAstV genera). Genome sequencing technologies have led to an exponential increase in the identification of emergent astroviruses and a large and growing dataset of genomes awaiting taxonomic assignment. These assemblies show evidence of cross-species transmission, recombination, and high mutation rates. These complexities challenge species-level classifications that prioritize host species. Recent advances in alignment-free genome sequence analyses with supervised machine learning and unsupervised clustering algorithms provide ultrafast, reliable, and sensitive genome taxonomic assignment. In this study, the Three-Pronged Classification Method (3PCM) was applied to 636 mamastrovirus genomes and 153 HAstV’s were separated from non-human MAstV’s (with accuracies of 99.36% and 80.88% in supervised and unsupervised approaches, respectively). Subclustering analysis generated three subclusters within an HAstV subgenera. Subcluster one contained bastrovirus (BAstV) genomes isolated from human feces. BAstV’s are also found in swine, rat, and bat, suggestive of cross-species transmission potential of viruses in this subcluster. Subcluster two predominantly contained VA1-4 and MLB HAstV strains from fecal and brain-derived samples associated with neurotropic effects as corroborated by previous evidence of tissue tropisms of these strain types (VA and MLB clades). Subcluster three contained fecal-derived genomes with classical HAstV serotypes 1-8 associated with gastrointestinal effects. Thus, 3PCM machine learning enhanced by principal component analysis achieved suggested assignments of previously unclassified astroviruses to subgenera HAstV with intraspecific classification associated with three HAstV clades.
