Repeated Nanopore Readout of the Same Peptide Moves Single-Molecule Protein Sequencing Closer to Reality

DNA sequencing transformed the life sciences, but transferring similar capabilities to proteins is far more difficult. Proteins not only contain 20 common amino acids but also undergo post-translational modifications such as phosphorylation. If these differences could be read directly at the single-molecule scale, researchers might be able to detect rare protein variants obscured by conventional bulk-average analyses. A team at Nanjing University has now published a nanopore method in *Nature* that takes a step toward this goal.

The technology is called transient pore analyte looping (tPAL). The researchers engineered an MspA nanopore derived from *Mycobacterium smegmatis*, installing a nickel ion-binding adaptor structure inside the pore and immobilizing the peptide to be tested. The peptide’s N-terminus repeatedly enters the sensing region, generating multiple, comparable ionic-current signals instead of rapidly passing through the pore and leaving only a single reading.

The coordination between reading and cleavage is what actually produces sequence information. The team introduced a cholesterol-modified aminopeptidase, allowing the enzyme to remove one amino acid at a time from the N-terminus. After each amino acid is removed, the new N-terminus enters the nanopore’s sensing position, producing the next stage of current changes. By progressively shortening the same immobilized peptide in this way, the continuous signals can be converted into clues for inferring its sequence.

The study showed that tPAL can distinguish differences between peptides at single-amino-acid resolution, including single-site substitutions, post-translational modifications, and the insertion of non-natural amino acids. For a set of 20 peptides differing only in their N-terminal residue, the team collected signals at three voltages and extracted multiple features before using machine learning for classification. A quadratic support vector machine achieved a maximum accuracy of 98% in tenfold cross-validation. This result demonstrates that the signals are highly discriminative, but it remains a validation within a controlled dataset and cannot be directly equated with sequencing accuracy for unknown clinical samples.

The value of this design lies in using “rereading” to reduce the fluctuations common in single-molecule signals while retaining chemical information such as modifications and non-natural residues. If it can be extended to longer and more complex protein mixtures, it may eventually address some of mass spectrometry’s limitations in analyzing low-abundance molecules, protein isoforms, and cell-to-cell differences. It may also be used for the precise identification of disease biomarkers or therapeutic peptides.

At this stage, tPAL should not be regarded as a complete protein-sequencing platform capable of directly replacing mass spectrometry. The method requires peptides to be immobilized first and then sequentially cleaved from the N-terminus, and the readout itself consumes the molecules being tested. Publicly available information from the paper has not yet shown that it can reconstruct proteomes at scale from predefined complex biological samples with sufficient throughput, coverage, and error control. The machine-learning model’s ability to generalize across different nanopores, batches, and sample conditions also awaits validation with independent data.

To move from proof of concept to a practical tool, the next steps will be to increase parallelization, build a more comprehensive signal library for amino acids and modifications, and address enzymatic cleavage efficiency and signal loss during long-sequence reads. The research team has filed patent applications covering the tPAL method and its applications. The most critical scientific test will be whether it can move beyond carefully designed peptides and reliably read unknown molecules in real cells and clinical samples.