A peptide written as letters is a compressed structural claim, and the compression is lossy in specific, enumerable ways. The string fixes which side chains are present and in what order. It does not fix configuration at any stereocentre, the state of either terminus, the location of any covalent modification, or which direction the backbone runs. The IUPAC-IUB recommendations that define both residue alphabets supply defaults for some of this, but a default is only safe when the reader knows it is operating [1]. Once a peptide carries chemistry outside the coded residues the deficit stops being a matter of defaults: the presence of non-natural chemical modifications makes representation by sequence alone impossible, which is the premise the HELM notation was built on [2].
Two alphabets with different jobs
The three-letter and one-letter systems are not interchangeable shorthands for the same content. The recommendations state that the one-letter system is less easily understood by those unfamiliar with it, that it should not be used in simple text or in reporting experimental details of sequence determination, and that it is intended for comparisons of long sequences in tables and lists and other uses where brevity is important [1]. The three-letter system is where substituents, non-coded residues and backbone alterations are actually writable.
The one-letter set also has analytical ambiguity built into it. B denotes aspartic acid or asparagine where the two have not been distinguished; Z denotes glutamic acid or glutamine, and explicitly also covers substances such as 4-carboxyglutamic acid and 5-oxoproline that yield glutamic acid on acid hydrolysis of peptides; X means the identity of the residue is undetermined or that the residue is atypical [1]. A B is a statement about the hydrolysis, not about the molecule. Several letters were left out of the original assignment for legibility rather than chemistry: U was avoided because it is easily confused with V in handwritten material, O because it is confusable with G, Q, C and D in imperfect print-outs and also with zero, and J because it is absent from several languages [1]. Those are typographic conventions rather than facts about molecules. A bare string does not record which revision of the alphabet produced it.
The recommendations further define punctuation for partly determined sequences: parentheses around a region whose composition is known but whose order is not, dots between the letters inside those parentheses where a residue can be positioned with confidence by homology and commas where its position is arbitrary for lack of evidence, and a slash separating residues that have not been shown experimentally to be connected or marking an end where termination has not been demonstrated [1]. A plain FASTA line carries none of this. Stripping it silently converts a hedged result into an apparently determined one.
Direction, and what an unstated one leaves open
The convention is explicit: the letter at the left-hand end is the residue carrying the free amino group, and the letter at the right-hand end is the residue carrying the free carboxyl [1]. Peptide names are constructed the same way, which is why glycylalanine and alanylglycine denote two different dipeptides [1]. Reversal is not a symmetry operation. A string whose direction is unstated therefore describes two candidate compounds of identical composition and identical mass, differing in every backbone contact they can make. Nothing in a sequence file forces the direction to be declared, and nothing in a mass measurement recovers it.
Configuration, and the two ways to establish it
Residue symbols in a sequence denote the L configuration for chiral amino acids unless otherwise indicated, and a D residue is shown by inserting a D before the symbol [1]. The prefix DL signifies a racemic mixture and so should not appear in the designation of a peptide with more than one chiral residue, because coupling a DL-amino acid to a chiral peptide gives diastereoisomeric products whose ratio may depend on reaction conditions and will not in general be unity; the prefix ambo exists for a mixture at one named centre [1]. Writing DL where ambo is meant asserts a racemate that is not what is in the vial.
Configuration is also the property a mass measurement cannot see, and the two routine ways to establish it are chemically different. Marfey's analysis relies on complete hydrolysis of the peptide followed by reaction of the resulting amino acid pool with a chiral derivatising agent, turning enantiomers into diastereomeric pairs that conventional reversed-phase HPLC can resolve [3]. It reports the pool and not the position, because hydrolysis discards the sequence. It also has a residue-specific limit: isoleucine and threonine possess a second chiral centre at Cβ and occur in natural non-ribosomal peptides as their allo diastereomers, and discrimination between the two possible epimers can be challenging for lack of chromatographic resolution. Pérez-Victoria and colleagues addressed this by trapping the HPLC peaks and comparing the NMR spectra of the Marfey derivatives against those of standards, which assigns the absolute configuration at the second centre; they demonstrated it on two model cyclic peptides bearing L-Ile and L-Thr [3]. An assignment that names only the Cα configuration has left the Cβ centre open.
One route to localising a D residue without destroying the chain is a separation that responds to shape rather than mass. Jia and colleagues separated D/L peptide epimers by online reversed-phase LC, fragmented them by collision-induced dissociation, and analysed the fragments by ion mobility: epimeric fragment ions show conformational differences and therefore different mobilities, and the arrival-time shift between them identifies which fragment carries the substitution [4]. They demonstrated the approach on [D-Trp]-melanocyte-stimulating hormone, [D-Ala]-deltorphin and [D-Phe]-achatin-I against their all-L counterparts, and used it to place the L- to D-Phe isomerisation at the third residue from the N-terminus in the 8.5 kDa crustacean hyperglycaemic hormones isolated from Homarus americanus [4].
Isomers that survive the mass measurement
Leucine and isoleucine share an elemental composition, so mass alone does not separate them; radical-driven side-chain cleavage does. Lebedev and colleagues used multistage dissociation on an Orbitrap Fusion — electron-transfer dissociation to produce and isolate odd-electron z• ions, then higher-energy collisional activation to drive radical-site fragmentation into w ions characteristic of the isomeric residues. The experiments succeeded in all 22 Leu/Ile cases across six natural non-tryptic peptides of 15 to 37 residues isolated from the secretion of Rana ridibunda, each peptide carrying between one and seven of the isomeric residues, with the targeted w ions appearing as the most intense peaks in the spectra [5].
Aspartate carries an isomer problem of the same shape. Isoaspartate has the Cβ incorporated into the backbone, lengthening the chain by one methylene unit while leaving the elemental composition untouched. Electron capture dissociation separates the two: model peptides with isoaspartyl residues gave cn+58 and zl−n−57 fragment ions, where n is the position of the residue and l the peptide length, found only in the isoaspartyl spectra and attributed by the authors to cleavage of the Cα–Cβ backbone bond that splits the residue between the two fragments; the aspartyl peptides instead showed a neutral loss of the aspartic acid side chain from the charge-reduced species [6]. A D in a sequence string is a two-molecule claim until that experiment, or an equivalent, has been run.
Termini and residues substituted on nitrogen
Terminal amidation and acetylation are chemistry, and the one-letter alphabet has no slot for either. The three-letter system does. Substituent groups are written as symbols attached to the sequence — acetyl as Ac-, a methyl ester as -OMe — so that glycine methyl ester is Gly-OMe and thyroliberin is written Glp-His-Pro-NH2, with both the pyroglutamyl residue and the C-terminal amide explicit [1]. Nitrogen substitution is handled the same way: glycylsarcosine and glycyl-N-acetylglycine are written as residue symbols carrying the substitution, so backbone N-methylation is representable in the three-letter system and simply absent from a one-letter string [1]. The recommendations also warn that where a modified residue is given a constructed symbol rather than a name, the symbol should be explained to avoid confusion [1]. A peptide reported as bare letters asserts, by default, a free α-amine, a free α-carboxyl and an unmethylated amide at every position.
Cyclic, and the ring atoms nobody names
"Cyclic" is not one structural class. Homodetic cyclic peptides are those whose ring consists solely of amino-acid residues in eupeptide linkage — the amide bonds formed between C-1 of one residue and N-2 of the next — and they are written with the sequence in parentheses preceded by cyclo, as in gramicidin S, cyclo(-Val-Orn-Leu-D-Phe-Pro-Val-Orn-Leu-D-Phe-Pro-) [1]. Heterodetic cyclic peptides consist only of amino-acid residues as well, but one or more of the ring-closing linkages is an isopeptide, disulfide, ester or other bond rather than a eupeptide bond [1]. Depsipeptides, oligomers formed from amino acids together with other bifunctional acids and usually hydroxy acids, are often cyclic, and any special symbols used for the hydroxy acids have to be defined [1]. Backbone replacement of the -CO-NH- group is a separate notation again, a psi followed by the replacing group in parentheses between the two residue symbols, so a reversed amide is Ala-psi(NH-CO)-Ala [1].
The word "cyclic" attached to a linear letter string therefore specifies nothing that matters: not the ring size, not the two atoms joined, not the bond type. A head-to-tail lactam, a side-chain lactam between an Asp or Glu carboxyl and a Lys or Orn amine, a lactone to a Ser or Thr hydroxyl, and a disulfide-closed macrocycle are four different compounds compatible with one sequence. The cyclo(-...-) form at least fixes a head-to-tail ring; every other case has to name both attachment atoms.
Disulfides: a count is not a connectivity
Cysteine positions live in the sequence; pairings do not. Peptides and proteins carrying high numbers of cysteines can adopt various three-dimensional structures depending on their disulfide connectivities, and verifying which isomer is in hand is hard. Heimer and colleagues synthesised all 15 possible isomers of the 22-mer conopeptide μ-PIIIA, which carries three disulfides, and applied 2D NMR spectroscopy and MS/MS to elucidate the structure [7]. Their operational conclusion is the one to carry forward: analysis procedures involving comprehensive combinations of conventional methods are required for unambiguous assignment of disulfides in cysteine-rich peptides, and standard compounds are crucially needed for that structural analysis [7]. A record giving six cysteines and the phrase "three disulfide bonds" has named a set of fifteen candidates, not a molecule.
Controlled vocabularies: what a modification term buys
"Oxidised", "acetylated" and "cyclised" are not chemical statements. Two public resources fix that from opposite ends. Unimod is a database of protein modifications for mass spectrometry, holding accurate and verifiable values, derived from elemental compositions, for the mass differences introduced by both natural and artificial modifications [8]. Its unit is a delta — what a search engine and a fragment ladder consume. RESID catalogues the modified residue instead: pre-, co- and post-translational modifications and cross-links, with systematic and alternate names, atomic formulas and masses, the enzymatic activities that generate them, keywords, literature citations, Gene Ontology cross-references, sequence-database feature-table annotations, structure diagrams and molecular models [9]. RESID is used to standardise and enhance modification descriptions in Swiss-Prot feature tables and supplies the table of equivalent feature annotations used in the UniProt project [9]. Quoting an accession from either replaces an adjective with a defined structure and a defined mass.
ProForma: notation a parser can act on
ProForma was proposed by the Consortium for Top-Down Proteomics as a standardised way to write the sequence of a fully characterised proteoform — the amino acid sequence in standard one-letter notation, with modifications or unidentified mass shifts specified within brackets following certain amino acids, under seven rules, unambiguous and human-readable while remaining easy for software to parse and write [10]. ProForma 2.0, developed by the HUPO Proteomics Standards Initiative in collaboration with members of the same consortium, aims to unify the representation of proteoforms and peptidoforms, and supports the use cases of bottom-up and middle- and top-down workflows. It encodes modifications in a specified or ambiguous location, designated by mass shifts, chemical formulas or controlled vocabulary terms, including natural and chemical cross-links and atomic isotopes, with notational conventions based on public controlled vocabularies and ontologies — PSI-MOD, Unimod, RESID, XL-MOD and GNO [11].
The syntax is worth seeing because it makes the failure modes visible. A located, identified modification is written EM[Oxidation]EVEES[Phospho]PEK. The same peptide with the chemistry unknown but the mass measured is EM[+15.9949]EVEES[+79.9663]PEK [11]. A modification known to be present but not localised takes a leading question mark, [Phospho]?EMEVTSESPEK. A localisation spread over candidate sites with scores is EMEVT[#g1(0.01)]S[#g1(0.09)]ES[Phospho#g1(0.90)]PEK, and a modification confined to a stretch rather than a residue is written against a parenthesised range, PROT(ESFRMS)[+19.0523]ISK. A disulfide is a labelled cross-link tying two positions together, EVTSEKC[MOD:00034#XL1]LEMSC[#XL1]EFD [11]. Each of those strings encodes a different state of knowledge, and none of them collapses into any other.
Where free text becomes unrecoverable
An unlocated modification is a legitimate result, and there are established ways to say how unlocated it is. The Ascore is a probability-based score that measures the probability of correct phosphorylation site localisation from the presence and intensity of site-determining ions in MS/MS spectra; the methods it belongs to, applied in a fully automated pipeline to nocodazole-arrested HeLa cell lysate with data-set error measured by the target-decoy approach, returned 1,761 non-redundant phosphorylation sites from 491 proteins at a 1.3% peptide false-positive rate [12]. Mass-tolerant searching extends the same logic to modifications whose chemistry is unknown: an ultra-tolerant Sequest search permitting modifications of unknown masses up to ±500 Da, run over a proteome-wide HEK293 data set of 9,513 proteins and 396,736 peptides, matched an additional 184,000 modified peptides falling into 523 distinct mass bins, localised all the unknown modification masses to specific regions within a peptide, and assigned known modifications to the correct amino acid with frequencies above 90%; the authors concluded that at least one third of unassigned spectra arise from peptides with substoichiometric modifications [13].
Both of those outcomes are reportable, and both have a ProForma form. What has no form and no route back is a free-text phrase that fails to separate measured from inferred from assumed. "Phosphorylated" with no residue index cannot be distinguished from a localisation that was performed and then not written down. "Cyclic" with no ring atoms cannot be reconstructed, because the sequence is consistent with every ring. "Contains D-amino acids" with no positions is beyond recovery from the deposited record, since composition, mass and letter string are all unchanged by the substitution. In each case the information existed at the bench and was destroyed at the keyboard.
What has to travel with the string
For sequence collections, the PSI extended FASTA format keeps the familiar format and adds a uniform mechanism for encoding substantially more metadata about the sequence collection and about individual entries, including known sequence variants, post-translational modifications and proteoforms. It is very nearly backward compatible, so existing FASTA parsers need little or no change to read PEFF files as FASTA while ignoring the extra capabilities, and it is defined by a full specification document, controlled vocabulary terms, example files, software libraries and a validator [14]. For a synthetic peptide whose structure is fully defined, line notations built for the job carry connectivity that a residue string cannot: HELM represents such biomolecules hierarchically rather than atom by atom, which keeps chemically modified and non-natural moieties tractable for registration, storage and analysis [2], and BILN is a human-readable line notation for complex peptides [15].
The practical minimum for a peptide record follows from the failure modes above. State the direction rather than leaning on the default. Give configuration at every stereocentre, and name the method that established it. Give both terminal states explicitly. Give every modification as a vocabulary term or a formula, with a position and the evidence class for that position. Name the two atoms of any ring. Name the cysteine pairings. Whatever is left to a default will be read downstream as measured.