Rational Engineering of the Anthrax Toxin Nanopore Interface for Orthogonal Peptide Classification

The development of high-performance functional biointerfaces requires precise control over molecular recognition mechanisms at the nanoscale. A fundamental limitation in current biological nanopore materials is that wild-type (WT) channels often lack the specific physicochemical "grip" required to resolve chemically similar analytes. Here, we demonstrate that rationally engineering a key active site, called the φ-clamp, in the anthrax toxin protective antigen (PA) nanopore creates a tunable biointerface with orthogonal selectivity. By mutating the φ-clamp constriction (F427A), we generated a variant that─despite being a defective large protein biological translocase─functions as a superior material for molecular recognition, achieving ∼93% classification accuracy for peptides that confound the WT pore. We integrated the WT and F427A variants into a multistage machine learning ensemble (XGBoost) to leverage the complementary physicochemical selectivity of each pore─WT for aromatics and F427A for small polar residues─achieving a balanced F1-score of >0.91. This work proposes a design principle for multiplexed biomaterials, demonstrating that attenuating native transport function via rational engineering can enhance the specificity of the biointerface for targeted analytical applications.