The analyser never reports a mass. It reports m/z, and for a peptide sprayed from acidified solvent that ratio carries an unknown integer in its denominator. Recovering a mass means deciding what that integer is at every peak worth keeping, and that decision is an inference made by software from a pattern that salts, solvent additives and the peptide's own count of basic sites all reshape. What emerges looks like a measurement on a mass axis. Part of it is a model, and the model has characteristic ways of producing peaks that correspond to nothing in the vial.

The charge-state envelope is the whole of the evidence

Electrospray does not place one charge on a peptide. The distinguishing features of electrospray spectra for large molecules are coherent sequences of peaks whose component ions are multiply charged, the ions of each peak differing by one charge from those of adjacent neighbors in the sequence [1]. That one-charge spacing is the entire basis of charge deconvolution: two adjacent members of a sequence give two equations in two unknowns, and every further member is a redundant check on the answer. A mass derived from six charge states and a mass derived from two are not the same kind of number, even when they agree to the same decimal place.

How many members you get is not something you set. The top of the envelope is governed by gas-phase proton transfer reactivity. Applied to a published series of arginine-containing peptide ions, a model of that reactivity account[ed] exactly for the maximum charge state for 7 of the 10 peptides and was off by one charge for the remaining 3, which the authors read as further evidence that the maximum charge state of electrosprayed ions is determined by their gas-phase proton transfer reactivity [2]. For a research peptide of a dozen residues carrying one or two basic sites, that ceiling is low. You may be deconvolving two peaks, and two peaks cannot disagree with each other.

The envelope is also movable by solution chemistry, which is worth establishing before concluding that a peptide ionises badly. Working on cyanobacterial cyclic heptapeptides, Yuan and colleagues found that for analytes already giving abundant [M+2H]2+ ions, the addition of nitrogen-containing bases to the aqueous sample solution is effective to obtain an increased intensity of [M + H]+ ions, whereas the addition of Lewis acids containing nitrogen can produce increased abundances of [M + 2H]2+ ions for those giving weak doubly protonated ions [3]. A charge-state distribution is a property of the experiment, not only of the molecule, so two laboratories reporting different envelopes for the same peptide are not necessarily contradicting each other.

Monoisotopic, average, and the peak that is not there

Two masses can be quoted for the same peptide. The monoisotopic mass sums the lightest stable isotope of every element in the formula; the average mass weights each element by natural isotopic abundance. Which of the two you can honestly report is decided by atom count, because the large number of atoms incorporated in these molecules results in a low probability for the all-monoisotopic species, and that low probability produces the potential to misassign the number of heavy isotopes in a specific peak and make a mass error of ±1 Da, although the certainty of the measurement beyond the decimal place is greater than 0.1 Da [4]. Run that mechanism backwards and the crossover falls out. At small atom counts the all-light isotopologue is still probable enough to sit at the front of an observable cluster, and monoisotopic is the number to quote; the crossover to average mass is not a mass threshold to memorise but the point at which that peak stops being observable.

That failure mode deserves recognition on sight, because it does not look like a failure. The spectrum is internally excellent, the peak centroid is reproducible to well under 0.1 Da, and the reported mass is a whole dalton wrong because the wrong member of the cluster was called monoisotopic. The remedy in the original work is not more resolution but a statistical comparison of the measured isotopic distribution against the distribution computed for a model molecule of the same average molecular mass, which the authors show permits assignment of the monoisotopic mass even in cases where the monoisotopic peak is absent from the spectrum [4]. Every modern isotope-fitting routine does some version of this, which means a monoisotopic mass for a large species is a fitted quantity, not a peak you can point at.

What the algorithms do, and the peaks they invent

Deconvolution software takes the m/z spectrum and returns a zero-charge mass spectrum. The maximum-entropy approach that established the genre produced those zero-charge spectra along with probabilistic quantification so that the reliability of features in the spectrum can be ascertained, and because maximum entropy stays faithful to the data the results tend to have improved resolution and signal-to-noise ratio; the demonstration resolved two haemoglobin β-globins separated by 12 Da at m/z 15 867 (0.08%) and a second pair separated by 9 Da (0.06%) that had not previously been resolved [5]. Those are intact-protein numbers, not peptide numbers, but they set the expectation that deconvolution improves apparent resolution — and improved apparent resolution is exactly the property that makes a deconvolved artefact convincing.

The Bayesian generation states the problem more explicitly: Interpretation of mass spectra is challenging because they report a ratio of two physical quantities, mass and charge, which may each have multiple components that overlap in m/z, and the solution is that incorporation of the charge-state distribution in the Bayesian prior probabilities provides separation of the m/z spectrum into its physical mass and charge components [6]. Read that as an instruction about your own responsibilities. You supply a prior describing what a plausible charge-state distribution looks like, and the algorithm will find one.

Give it too much room and it will find several. Charge-deconvolution algorithms, when applied over a wide input m/z or broad target mass range, can produce artifacts, such as false masses at one-half or one-third of the correct mass, and a maximum entropy term in the objective function of the most commonly used algorithm favors a deconvolved spectrum with many peaks over one with fewer peaks [7]. These are harmonics, and their arithmetic is simple: a series read at half the true charge yields half the true mass, a third of the charge yields a third of the mass. They are also the easiest artefact to design out, by setting the target mass window and charge range from what the synthesis and the chromatography allow rather than from the widest values the software will accept.

Newer algorithms have their own version of the same weakness. The UniDec algorithm can produce artifacts depending on the quality and complexity of the data; applying a SoftMax function to the charge-state distribution of each data point pushes the algorithm towards assigning each data point to one primary charge state and significantly reduces deconvolution artifacts, even for data with overlapping charge states [8]. The general lesson is that artefact suppression is a tunable parameter, so a deconvolved spectrum reported without its parameters is not reproducible.

Sodium, potassium and trifluoroacetate: the distortion you brought with you

Adducts are not a contamination event; for a synthetic peptide they are the default state of the sample. Most synthesized peptides are nowadays produced using solid-phase procedures. Due to cleavage and purification conditions, they are mainly obtained in the presence of trifluoroacetic acid (TFA) and, for cationic peptides, as trifluoroacetate (TF-acetate) salts [9]. TFA is present before any mobile phase is chosen, and exchanging it out takes deliberate work: reverse-phase HPLC and ion-exchange resin gave partial to almost complete exchange of the tightly bound counter-ion on a dicationic octapeptide, while only a deprotonation/reprotonation cycle removed it completely [9].

What TFA then does to the spray is mechanistically specific. Its signal suppressing effects are due to the combined effect of ion-pairing and surface tension modifications, and post-column addition of propionic acid–2-propanol (75:25, v/v) in a 1:2 proportion with the mobile phase counteracts them, giving a 10-100 x improvement of the signal-to-noise ratio in that peptide-mapping configuration [10]. Suppression and adduction are two faces of the same solution chemistry, and the anion decides which face you see: for multiply charged protein ions, the extent of sodium and acid molecule adduction is inversely related and depends strongly on the proton affinity (PA) of the anion, with extensive sodium adduction for anions of PA above about 300 kcal mol−1 and extensive acid-molecule adduction for anions below 315 kcal mol−1, so salts with anions that have low PA values may be useful solution-phase additives to minimize nonspecific metal ion adduction [11]. That measurement was made on proteins; treat it as a rationale for choosing counter-ions rather than as a peptide result.

The peptide-scale consequence is that the protonated species need not be the base peak at all. Those microcystins possessing no arginine residue always give sodium adduct ions [M + Na]+ as the base peak, and those adducts are difficult to fragment via low energy collision-induced dissociation to yield structurally informative products until oxalic acid is added to restore [M+H]+ [3]. A peptide with no strongly basic residue can therefore hand you an envelope whose dominant series is a sodiated one, and a deconvolution told to expect a single mass will either fold that series into the parent or emit it as a satellite.

Sodium replacing a proton shifts the neutral mass by about 22 Da and potassium by about 38 Da, and those offsets are constant in mass while the charge states are spaced in m/z. That gives the diagnostic: a satellite sitting a fixed +22 or +38 from the parent in the deconvolved spectrum, reproducing the parent's charge-state envelope shape, is salt. A satellite at an offset that does not correspond to an alkali metal, and that survives desalting, is chemistry. The distinction is worth making before a mass difference is written down as a modification.

Resolution decides what the isotope pattern can tell you

Resolution matters because the interferences that defeat intact-mass work are millidalton-scale, not dalton-scale. Deamidation is the standard example: mass spectral analysis of deamidated peptides is complicated and often misassigned due to overlapping (13)C peak of the amidated form with the deamidated monoisotopic peak; these two peaks are only separated by 19.34 mDa [12]. In a global proteomics experiment on mouse brain tissue lysate analysed on a quadrupole-Orbitrap instrument, precursor searches run with greater than 5 ppm mass tolerance resulted in the misassignment of deamidated peptides, while peptides that were identified with a mass measurement accuracy of ±5 ppm were correctly assigned [12]. The tolerance you set in the search is doing as much work as the analyser.

Where the species genuinely overlap, no amount of peak picking separates them, and the measurement has to become a fit. Because the isotopic envelopes of mass-shifted peptides are normally not resolved, quantification cannot be performed by simple integration of extracted ion chromatograms; fitting the observed envelope with predicted isotopic envelope mixture models handled it, and on synthetic peptides mixed with their deamidated forms at ratios from 1:0 to 20:1 the method gave a correlation of R2 = 0.96 between measured and known percentage deamidation [13].

Resolution is not, however, the only route to a mass-degenerate assignment, and it is a mistake to assume it is. Trimethylation and acetylation differ by 42.0470 versus 42.0106 Da, and high-resolution analysers have been regarded as essential precisely because that difference is resolvable there and not on low-resolution instruments. Comparing a low-resolution ion trap with a high-resolution Orbitrap on histone peptides, the main difference was exactly that resolving power — but using heavy isotope labeled synthetic peptide standards and retention time information enables confident assignment of these modifications and comparable quantification between the instruments, so lower-resolution instruments can be used for that analysis when those orthogonal handles are in place [14]. Resolution buys you the assignment directly; retention time and authentic standards can buy the same assignment indirectly.

What an accurate mass licenses, and what it does not

A mass constrains composition, and the constraint has a published ceiling. At a mass accuracy of about ±1 ppm, the mass limit for revealing the unique elemental composition of a peptide was found to be 700–800 Da, with the possibility to extend this range as the mass accuracy improves [15]. The corresponding limit for determining amino acid composition is roughly 500–600 Da and, in the general case, cannot be overcome by instrumental or methodological improvement; separately, ±1 ppm was found to exclude 99% of peptides sharing the same nominal mass but differing in elemental and amino acid composition [15]. Those are three different numbers answering three different questions, and they are not interchangeable: exclusion of same-nominal-mass alternatives is a much weaker claim than determination of a unique formula.

Above those limits an intact mass is a filter, not an identification, and the residual ambiguity is structural rather than compositional. Isomers are the clean case. Leucine and isoleucine are identical in formula and therefore in mass at any resolving power, and separating them requires fragmentation chemistry: multistage electron transfer dissociation followed by higher-energy collisional activation generates w-ions characteristic of the isomeric residue, an approach that succeeded in all 22 Leu/Ile positions across six natural non-tryptic peptides of 15 to 37 residues [16]. Sequence permutations, D- and L-residues, and the trimethyl/acetyl pair below adequate resolving power are all mass-degenerate in the same way. An intact mass that matches to 1 ppm tells you the composition is consistent; it does not tell you the atoms are arranged as intended. That is the boundary at which this measurement hands off to the fragmentation-based workflow described in the guide on mass-spectrometric peptide identity.

Reporting a deconvolved mass so someone else can check it

  • State which charge states contributed. A mass from two adjacent states carries no internal redundancy; a mass from a full envelope does.
  • State the algorithm, the mass window, the charge range and any artefact-suppression setting, since these determine whether harmonics at half and a third of the mass could have appeared at all [7, 8].
  • Say whether the number is monoisotopic or average, and for a large species say whether the monoisotopic value was observed or fitted from the envelope [4].
  • Map every deconvolved peak back into the raw m/z domain. A real mass has visible peaks at the claimed charge states; a harmonic does not.
  • Record the counter-ion the material arrived with and the anion present in the spray, since both govern how much adduction the envelope carries [9, 11]. Satellites at about +22 and +38 are then evidence about sample handling rather than about the peptide.

Deconvolution in native conditions is a harder version of the same problem rather than a different one, and the difficulties are enumerated in the parsimonious-deconvolution work: native spectra pose special challenges due to salt and small molecule adducts, multimers, wide mass ranges, and fewer and lower charge states [7]. Fewer charge states means less redundancy in exactly the sense above, and adducts and multimers give the algorithm more ways to be wrong, which is why the native mass spectrometry guide treats stoichiometry assignment as a separate argument rather than a readout. Under denaturing conditions with a well-populated envelope and a desalted sample, an intact mass is one of the most reliable numbers a peptide laboratory can produce — provided it is reported as what it is, which is the output of a fit.