METHODS: Samples were obtained from 172/192 children presenting to a site in rural India with acute encephalitis syndrome.

RESULTS: Using the reference VT ELISA, infection with Japanese encephalitis virus (JEV) was confirmed in 44 (26%) patients, with central nervous system infection confirmed in 27 of these; seven patients were dengue seropositive. Of the 121 remaining patients, 37 (31%) were JEV negative and 84 (69%) were JEV unknown because timing of the last sample tested was <10 day of illness or unknown. For patient classification with XCyton, using cerebrospinal fluid alone (the recommended sample), sensitivity was 77.8% (59.2-89.4) with specificity of 97.3% (90.6-99.2). For Panbio ELISA, using serum alone (the recommended sample), sensitivity was 72.5% (57.2-83.9) with specificity of 97.5% (92.8-99.1). Using all available samples for patient classification, sensitivity and specificity were 63.6% (95% CI: 48.9-76.2) and 98.4% (94.5-99.6), respectively, for XCyton ELISA and 75.0% (59.3-85.4) and 97.7% (93.3-99.2) for Panbio ELISA.

METHODOLOGY/PRINCIPAL FINDINGS: Using in silico tools, we designed and expressed four novel P. knowlesi protein products to address the distinct lack of suitable serosurveillance tools: PkSERA3 antigens 1 and 2, PkSSP2/TRAP and PkTSERA2 antigen 1. Antibody prevalence to these antigens was determined by ELISA for three time-points post-treatment from a hospital-based clinical treatment trial in Sabah, East Malaysia (n = 97 individuals; 241 total samples for all time points). Higher responses were observed for the PkSERA3 antigen 2 (67%, 65/97) across all time-points (day 0: 36.9% 34/92; day 7: 63.8% 46/72; day 28: 58.4% 45/77) with significant differences between the clinical cases and controls (n = 55, mean plus 3 SD) (day 0 p<0.0001; day 7 p<0.0001; day 28 p<0.0001). Using boosted regression trees, we developed models to classify P. knowlesi exposure (cross-validated AUC 88.9%; IQR 86.1-91.3%) and identified the most predictive antibody responses.