Predicting disease severity from an intra-host viral population of dengue virus serotype 2 in primary infection by machine learning

Dengue virus, a positive-sense single-stranded RNA virus, continuously threatens human health, especially in the tropical and subtropical regions worldwide. Although several criteria for evaluation of severe dengue have been recently established, the ability to prognose the risk of severe outcomes for dengue patient remains limited. Mutant spectra of RNA viruses, including single nucleotide variants (SNVs) and defective virus genomes (DVGs), have recently been reported to contribute to viral virulence and growth. Here, we aimed to determine the potency of viral mutant spectra in dengue patients with primary infection that progresses into severe dengue. We utilized deep sequencing to directly define the SNVs and DVGs of dengue virus in sera of dengue patients and analyzed associations between the mutant spectra and severe dengue. Among the detected SNVs and DVGs, 9 SNVs and 1 DVG exhibited statistically significant differences between patients with dengue fever and those with severe dengue. By utilizing the selected SNVs and DVG as prognostic features, the machine learning model showed high average discrimination with a high value of area under the receiver operating characteristic curve for severe dengue prognosis. In terms of the model, the elevation of detection times of DVG that had a junction in the E protein region (nucleotide positions of the junction: between 969 and 1022) and the frequency of SNVs at E (nucleotide position 995 and 2216), NS2A (nucleotide position 4105), NS3 (nucleotide position 4536, 4606), and NS5 protein (nucleotide position 7643 and 10067) increased the susceptibility of dengue patients for severe dengue. In summary, we analyzed SNVs/DVGs in dengue patients and developed prognostic models to predict dengue severity, using machine learning. Our results demonstrated that characterizing SNVs/DVGs among viral RNA genomes, combined with the machine learning model, may provide a potential prognosis method to triage dengue patients during dengue outbreaks; however, the model requires further validation using external cohorts in future studies. Additionally, profiles of the SNVs and DVGs that are associated with severe dengue will require further investigation to expand our understanding of mutant spectra of the virus, thus contributing to dengue pathogenesis in the future.