A single engineered protein pore can now read peptides one building block at a time. Paired with a machine-learning decoder, it can rebuild a peptide's sequence from its broken-up fragments. The work, published June 15 in Nature Nanotechnology ↗, brings to proteins the trick that made cheap, portable DNA reading possible. Thread a molecule through a tiny hole. Watch the electrical current dip as each unit passes. Decode the pattern of dips back into the sequence.
DNA gave that trick up more than a decade ago, and Oxford Nanopore Technologies turned it into a USB-sized sequencer. Proteins have not. The reason is chemistry. DNA is built from four bases, all carrying the same negative charge that an electric field uses to pull the strand through a pore in order. Peptides are built from 20 amino acids with different sizes, charges, and shapes, and no built-in handle to drag them through in single file. Telling 20 signals apart, instead of four, has kept protein nanopore sequencing stuck at the proof-of-idea stage for years.
What the pore actually read
The team used a modified version of a bacterial pore called MspA, from Mycobacterium smegmatis. They fitted it with a nickel anchor, hence the full name MspA-NTA-Ni, that briefly grabs analytes as they pass so the pore registers a clean signal. Under one set of conditions, the pore produced distinct current signatures for all 20 standard amino acids. It also read 4 chemically modified amino acids, 32 peptides, 6 modified peptides, 11 bioactive peptides, and 2 neoantigen peptides, the short tumor-specific fragments that cancer vaccines are designed around. A classifier trained on those signals sorted the analytes with a validation accuracy as high as 97.4 percent on the studied set.
Reading single amino acids is one thing. Reading a sequence is harder, and it is where the paper makes its real claim. The pore handled peptides up to 39 amino acids long, and the authors ran a sequencing proof of concept: take one reference peptide, chop it with two kinds of enzymes into overlapping pieces, read each piece through the pore, and let the machine-learning model stitch the fragment readings back into the original order. The reconstruction was sensitive enough to flag single-residue mutations, deletions, and post-translational modifications, the small chemical tags that switch a protein's behavior on and off and that mass spectrometry, the current workhorse, often struggles to localize.
Why a peptide reader matters
Mass spectrometry dominates proteomics today. It needs large, costly instruments and a lot of sample, and it infers sequences indirectly from mass fragments rather than reading them off in order. A nanopore that reads peptides directly, on the kind of small device that already exists for DNA, would change the economics of several jobs at once. Neoantigen discovery for personalized cancer vaccines depends on identifying exactly which tumor peptides are present. Detecting where a phosphate or sugar group sits on a protein is central to a lot of disease biology. And quality control on synthetic peptide drugs, confirming that the molecule in the vial is the sequence on the label, with no deletions or scrambled residues, is a recurring headache the pore's mutation sensitivity speaks to directly.
That last use case is the one closest to home for a platform like peptidemodel, where every card is anchored to a specific sequence. A cheap, direct way to confirm that a peptide is what it claims to be, residue by residue, is the kind of measurement that sequence-defined catalogs are built to consume.
The honest caveats
This is a proof of concept, not a product. The 97.4 percent accuracy is on a defined, studied set of analytes, not on an unknown protein pulled from a blood sample, and a 39-residue ceiling is far short of a full-length protein, most of which run to hundreds of amino acids. The hydrolysis-and-reassembly approach also depends on getting clean, overlapping fragments, which gets harder as molecules get longer and messier. What the paper establishes is that the central obstacle, telling the 20 amino acids apart through a single pore with enough fidelity to call mutations, is no longer the wall it was. The engineering from here is real, but it is engineering, not a missing principle.