# A dynamical anthrax toxin nanopore biosensor for high-fidelity single-peptide classification

## TL;DR

A new biosensor using a dynamic anthrax toxin nanopore can identify individual peptides with high accuracy, without needing to average many signals.

## Contribution

The use of a dynamic nanopore with moving parts enables single-molecule peptide classification with high accuracy.

## Key findings

- The anthrax toxin PA nanopore generates multi-state signals for individual peptides.
- A machine learning framework achieved ~91% accuracy in classifying single peptides.
- The method distinguishes peptides with identical mass but different atomic arrangements.

## Abstract

Nanopore sensing holds the potential to revolutionize proteomics, yet current methods often rely on ensemble aggregation, where thousands of events must be statistically aggregated or averaged to identify a protein or peptide signature. While effective for pure samples, this aggregation strategy fails in complex, heterogeneous mixtures where the identity of individual molecules must be determined in real-time. Here, we demonstrate high-fidelity classification of peptides from single, individual translocation events, eliminating the need for ensemble averaging. This sensitivity is achieved using the anthrax toxin protective antigen (PA) nanopore. Unlike static pores used more generally, the PA pore’s dynamic active-site clamps generate information-rich, multi-state signals. These clamps also enable the utility of high-affinity capture, permitting analysis at low nanomolar concentrations. We developed a machine learning framework that makes inferences on these dynamical multi-state signals and achieves ~91% accuracy on single events. This work establishes a framework for true single-molecule proteomics, enabling the resolution of complex mixtures that bulk aggregation methods cannot decipher.

Most nanopore sensors function like static molecular calipers, identifying molecules based solely on their size as they pass through a fixed channel. However, this approach often struggles to distinguish chemically similar molecules without averaging thousands of signals together. In this study, we introduce a fundamentally different approach by repurposing the anthrax toxin nanopore protein as a ‘dynamic’ biosensor. Unlike static pores, this biological machine possesses active moving parts that grab and interrogate molecules as they pass through. We show that these dynamic interactions generate complex, unique signal patterns—essentially a kinetic fingerprint—for every single molecule. Using a specialized machine learning framework, we successfully decoded these patterns to identify specific peptides with over 90% accuracy from just a single event, rather than an average. We could even distinguish peptides that were identical in mass but differed only in their atomic arrangement. This work demonstrates that by using a sensor that moves and adapts, we can achieve high-precision, single-molecule identification. This capability is a critical step toward future diagnostic tools capable of analyzing complex biological mixtures, such as blood, where traditional averaging methods fail to resolve individual components.