Deep learning–based identification of variant‑resilient broadly neutralizing antibodies to SARS‑CoV‑2

The integration of artificial intelligence with antibody research and development has paved new ways to respond more swiftly to viral outbreaks. However, current efforts mainly optimize existing antibodies or evaluate neutralizing activity against emerging SARS‑CoV‑2 variants, and rarely leverage multimodal information spanning both viral genomes and spike protein evolution. To address this gap, I will develop a deep learning‑based framework that integrates protein‑level features of SARS‑CoV‑2 spike and receptor‑binding domain with nucleic acid‑level information on viral mutations to predict broadly neutralizing antibodies against diverse and prospective variants. By systematically learning how sequence changes reshape antigenic landscapes and antibody escape, this AI‑based predictor aims to prioritize mutation‑resistant bnAb candidates and provide actionable guidance for rational vaccine and therapeutic antibody design.