Background: Virus phenotypic characteristics, such as virulence and transmissibility, are dictated by specific functional regions in viral proteins. This type of knowledge is largely buried in scientific publications in textual form. To make such information more easily searchable and analyzable, we have curated Sequence Features (SFs) from literature and database sources, and made them available through the Virus Pathogen Resource (ViPR; www.viprbrc.org), a US NIH/NIAID-funded, freely-available online Bioinformatics Resource Center supporting research on a broad range of human virus pathogens. These SFs not only provide useful protein sequence annotations, but also facilitate genotype-phenotype association studies. Methods and materials: We utilized the ViPR-curated SFs to explore the correlations between clinical manifestations of Dengue virus infection and Dengue genetic sequence variations to determine if specific SFs are associated with severe disease. Toward this aim, we used both statistical and machine learning methods to analyze a dataset of over 2,000 Dengue genomes isolated from patients with disease severity clinical metadata - mild (Dengue fever) or severe (Dengue hemorrhagic fever, Dengue shock syndrome). Results: First, Pearson's Chi-squared test was used to analyze 144 Dengue virus functional SFs and the sequence variations in these SFs (Variant Types, VTs) retrieved from ViPR. This analysis identified 10 significant SFs in NS2B, NS3, NS4A, NS4B and NS5 for Dengue 2 and one SF in NS5 for Dengue 3. Second, a severe disease classifier was built for Dengue 2 using Random Forest machine learning and 87 Dengue 2 functional SFs as classification features. Analyzing 563 sequences (75% for training and 25% for testing), 74% accuracy, 81% sensitivity and 64% specificity was achieved in the test set. Importantly, the top 10 SF features used by the severe disease classifier corresponded to the significant SFs identified by the Chi-squared test. These SFs include regions known to be involved in nuclear localization, ATPase and helicase activity, IFN antagonism, STAT degradation, and polymerase activity. Conclusion: In summary, ViPR SFVT data was used to study genotype-phenotype association in Dengue virus and identified SFs potentially associated with disease severity. This SFVT approach provides a way to analyze the effects of genetic variations on virus pathogenesis at a finer level.