Long-read protein sequencing

A protein is a chain of building blocks called amino acids. Reading a protein means finding their order, including chemical changes to those blocks. Can a small device read long, unfamiliar proteins quickly and accurately, without identifying them by matching a known sequence?[1] [2]

A protein is a chain of amino acids, chemical building blocks whose order forms its sequence. Reading an unfamiliar protein means identifying that order, including modified building blocks, rather than matching it to a known reference.

Can one microsystem do this while jointly meeting PROSE’s targets: proteins of at least 300 amino acids, at least 99% single-letter accuracy, at least 10 billion letters read per day, and an alphabet of at least 100 amino acids and modified forms? These are the program’s specific targets, not a definition of every useful sequencing method.[1] [2]

The PROSE program sets a combined target: chains at least 300 blocks long, 99% correct individual letters, 10 billion letters read per day, and 100 kinds of letter including modified forms. Those numbers must work together in an integrated device. Identifying short test peptides with prepared libraries is useful progress, but does not meet that target.[1] [2] [3]

EANPSeq identifies short model peptides by matching successively shortened fragments against libraries. Extending recognition to unfamiliar sequences and modifications, preparing complex samples and integrating long reads at the required rate remain further work.

Single-letter accuracy also differs from a perfect whole read. If each of 300 letters were independently correct with probability 0.99, the chance of no errors would be $0.99^{300}\approx0.049$. This is an illustration of the distinction, not a measurement of a device’s error pattern.[1] [2] [3]

DNA predicts a protein’s encoded sequence; it does not describe every chemical change to the resulting molecule. Direct reading could show molecular differences that a match to a familiar sequence misses. That would give researchers a fuller picture of what proteins are actually present in a sample.[1] [2] [3]

Reference identification asks which known sequence best explains a signal. Direct sequencing instead seeks the molecule’s sequence and chemical forms, including differences absent from that reference. Resolving those differences would make unfamiliar protein diversity observable. Possible applications still depend on reliable sample preparation and validated system performance.[1] [2] [3]