Identifying the targets of antibody responses during infection is important for designing vaccines, developing diagnostic and prognostic tools, and understanding pathogenesis. We developed a novel deep sequence-coupled biopanning approach capable of identifying the protein epitopes of antibodies present in human polyclonal serum. Here, we report the adaptation of this approach for the identification of pathogen-specific epitopes recognized by antibodies elicited during acute infection. As a proof-of-principle, we applied this approach to assessing antibodies to Dengue virus (DENV). Using a panel of sera from patients with acute secondary DENV infection, we panned a DENV antigen fragment library displayed on the surface of bacteriophage MS2 virus-like particles and characterized the population of affinity-selected peptide epitopes by deep sequence analysis. Although there was considerable variation in the responses of individuals, we found several epitopes within the Envelope glycoprotein and Non-Structural Protein 1 that were commonly enriched. This report establishes a novel approach for characterizing pathogen-specific antibody responses in human sera, and has future utility in identifying novel diagnostic and vaccine targets.

Emerging or re-emerging dengue virus (DENV) causes dengue fever epidemics globally. Current DENV serotypes are defined based on genetic clustering, while discrepancies are frequently observed between the genetic clustering and the antigenicity experiments. Rapid antigenicity determination of DENV mutants in high-throughput way is critical for vaccine selection and epidemic prevention during early outbreaks, where accurate prediction methods are seldom reported for DENV. Here, a highly accurate and efficient in-silico model was set up for DENV based on possible antigenicity-dominant positions (ADPs) of envelope (E) protein. Independent testing showed a high performance of our model with AUC-value of 0.937 and accuracy of 0.896 through quantitative Linear Regression (LR) model. More importantly, our model can successfully detect those cross-reactions between inter-serotype strains, while current genetic clustering failed. Prediction cluster of 1,143 historical strains showed new DENV clusters, and we proposed DENV2 should be further classified into two subgroups. Thus, the DENV serotyping may be re-considered antigenetically rather than genetically. As the first algorithm tailor-made for DENV antigenicity measurement based on mutated sequences, our model may provide fast-responding opportunity for the antigenicity surveillance on DENV variants and potential vaccine study.