We exploit the cross-reactivity of dengue (DENV) and Zika (ZIKV) virus polyclonal antibodies for nonstructural protein 1 (NS1) to construct a selective sensor that can detect yellow fever virus (YFV) NS1 in a manner similar to chemical olfaction. DENV and ZIKV antibodies were screened for their ability to bind to DENV, ZIKV, and YFV NS1 by enzyme linked immunosorbent assay (ELISA) and in pairs in paper immunoassays. A strategic arrangement of antibodies immobilized on paper and conjugated to different colored gold NPs was used to distinguish the three biomarkers. Machine learning of test area RGB values showed that with two spots, readout accuracies of 100% and 87% were obtained for both pure NS1 and DENV/YFV mixtures, respectively. Additional image preprocessing allowed differentiation between all four DENV serotypes with 92% accuracy. The technique was extended to hack a commercial DENV test to detect YFV and ZIKV by augmentation with DENV and ZIKV polyclonal antibodies.
Keywords: antibody cross-reactivity; machine learning; multicolor gold nanoparticles; multiplexed sensing; selective sensing.