Abstract
Finding therapeutic antibodies that bind multiple related targets with high affinity and favorable biophysical properties remains challenging and resource-intensive. For snakebite antivenoms, this challenge is critical as treatments must neutralize toxins across multiple snake species. We developed a pipeline combining high-throughput yeast screening, deep sequencing, and AlphaFold3 structure prediction to identify polyspecific variable domains of heavy chain–only antibodies (V
H
Hs) against long-chain α-neurotoxins. Multiplexed yeast display screening generated a dataset of diverse hits with varying binding specificities. AlphaFold3-generated V
H
H-toxin complex predictions enabled stringent high-precision structural triage of polyspecific V
H
Hs that bind conserved epitopes across multiple toxins, with a representative subset experimentally confirmed to block toxin binding to the acetylcholine receptor. These structural insights provided a starting point for computational optimization of affinity and soluble expression of the V
H
Hs, with experimental validation confirming that optimized V
H
H variants maintained broad binding specificity across toxins. This integrated approach supports structure-guided prioritization of polyspecific V
H
H hits, offering a high-precision filter that could reduce future reliance on extensive experimental specificity screening and providing a framework applicable to other therapeutic contexts, in which substantial antigen variation occurs and broad neutralization is essential.