A strategy for ionic current analysis of nanopore protein readouts
TL;DR
An integrated computational strategy for processing nanopore protein ionic current signals from synthetic protein constructs containing all 20 amino acids in controlled sequence contexts is developed, using a profile hidden Markov model to capture sequential structure of protein translocation.
Abstract
Background Nanopore-based protein reading holds promise for single-molecule proteomics, but analyzing the resulting ionic current signals presents substantial computational challenges. Unlike nucleic acids, peptide-associated ionic signals exhibit complex patterns with subtle transitions between amino acids that can be obscured by noise. We developed an integrated computational strategy for processing nanopore protein ionic current signals from synthetic protein constructs containing all 20 amino acids in controlled sequence contexts. We started with denoising to remove artifacts, followed by change point detection using variable-number and fixed-number segmentation methods. Results We employed dynamic time warping barycenter averaging to generate amino acid-specific reference templates. In parallel, we performed classification of amino acid regions using bidirectional LSTM networks trained on segment-level statistical features. Accuracy for 20-way amino acid classification remained limited, while charge-based binary classification reached 93.3% accuracy. Pairwise classification revealed that certain amino acids, particularly negatively charged residues, generate distinctive signatures. We used a profile hidden Markov model to capture sequential structure of protein translocation. Conclusion This approach provides computational strategies for automated segmentation, classification, and modeling of nanopore peptide translocation ionic current data.