We systematically tested five epitopes based on Larsen et al.'s deep mutational scanning data, and ran BoltzGen starting with the core EFNB2/3 discrimination site (490-494) that showed exceptional binding, then expanding to extended (488-496) and asymmetric (490-494,496) variants. The N306 glycosylation site (304-308) served as our secondary target. The asymmetric epitope strategy (490-494,496) was the one that produced our highest ranking binder with 0.62 IPSAE (soluble, without cysteines).