Computational Biology and Statistical Modeling Core (CORE B) (University of California, Berkeley) SUMMARY The Computational Biology and Statistical Modeling Core (Core B) will provide essential services to individual Projects and the Program Project (P01) as a whole by providing statistical support at each stage of research. Critically, Core B will enable the Projects to address key themes of the P01 by applying unifying computational biology, statistical, and machine learning approaches to study natural and vaccine-induced dengue humoral and cellular immunity. In Aim 1, Core B will conduct epidemiological analyses of the natural dengue virus (DENV) infection and dengue vaccine cohorts to inform the immunological studies proposed by Projects 1, 2, and 3. For the Nicaragua Pediatric Dengue Cohort Study, we will work closely with Core C to investigate dengue incidence before and after the introduction of Zika as well as how changing DENV transmission intensity affects dengue disease severity. For the Cebu Dengvaxia® cohort, we will estimate DENV infection and dengue disease incidence stratified by baseline DENV serostatus and vaccination history to support the immune correlates studies proposed by Project 2. We will also compare these two pediatric cohorts to understand how geography, DENV transmission intensity, ZIKV infection history, and serotype prevalence affect dengue disease. Phylogenetic and phylodynamic analyses will be conducted for all DENV isolates from both cohorts. In Aim 2, we will support each Project individually and conduct cross-Project analyses to identify immune markers that correlate with protection against symptomatic dengue and pathogenesis of severe dengue disease. This aim encompasses immune correlates of natural and vaccine-induced DENV immunity. We will work with each Project to design case-control studies to test how DENV-specific serum antibody, B cell, and T cell characteristics predict distinct clinical outcomes. Core B will analyze the multi-dimensional datasets produced by the Projects to classify clinical outcomes using straightforward machine learning methods such as generalized linear models, flexible approaches such as random forests, and methods that are robust to outliers such as support vector machines, all with regularization to reduce model complexity. In Aim 3, we will support the Projects in studying children who have experienced natural primary and secondary DENV infections to identify immune markers that predict maintenance anti-DENV immunity. We will use regression models to determine how antibody and helper T cell characteristics measured soon after infection predict both the magnitude and the durability of cross-reactive and type-specific antibody responses. We will then incorporate the predictive immune markers into linear and more flexible mixed-effects regression models to fit antibody dynamics following primary and secondary DENV infection. Parallel analyses will be conducted for baseline serone...