A peptide array is a surface carrying short synthetic sequences at known positions. Incubate it with an antibody preparation, wash it, detect what stayed with a labeled secondary reagent, and read one intensity per position. That is the whole measurement: an intensity map over a layout you designed. Everything after that — this stretch is an epitope, that residue is essential — is interpretation, and most of the discipline sits in the controls that decide whether it is allowed.
What an overlapping scan samples
The standard design cuts a protein sequence into peptides of fixed length, each starting a fixed number of residues after the one before. Three library types are described for antibody work on SPOT-synthesized arrays: sequence-derived scans of overlapping peptides, used to locate epitopes within the protein sequence; truncation libraries, used to identify the minimal peptide length required for antibody binding; and complete substitutional analyses, used to identify the key residues important for antibody binding [1].
The arithmetic of a scan is worth writing down, because it fixes what the experiment can and cannot see. For a protein of N residues scanned as k-mers stepping s residues at a time, the peptide count is floor((N minus k) divided by s) plus 1. A 300-residue protein scanned as 15-mers at single-residue offset gives 286 peptides; at a 3-residue offset it gives 96, about a third as many, the denominator in that comparison being the single-offset count of 286. Away from the termini, each residue then appears in up to five different peptides, because 15 divided by 3 is 5.
The offset also sets a guarantee. With length k and offset s, any contiguous stretch of k minus s plus 1 residues or fewer appears intact inside at least one peptide, again away from the termini — 13 residues for 15-mers at offset 3. A determinant longer than that may be split across two spots and appear whole in neither. Coarsening the offset is not a neutral saving: it trades peptide count against the longest stretch you are certain to have presented intact. None of this is a literature finding — it is counting, and it should be recomputed per design.
SPOT synthesis, and what a membrane costs you
SPOT synthesis dispenses droplets of activated amino acid onto a planar cellulose sheet; the wetted area defines a spot, and the chemistry proceeds as ordinary solid-phase synthesis on a flat, positionally addressable support. First presented in 1990 at the 21st European Peptide Symposium, the technique was characterized in a 2002 review as less impressively miniaturized than photolithographic array synthesis while fulfilling similar demands, with the advantages of a reliable and easy experimental procedure, inexpensive equipment needs, and highly flexible array and library formatting; it permits rapid, highly parallel synthesis of large numbers of peptides and peptide pools, including unnatural building blocks, and adapts easily to binding, enzymatic and cellular assays [2].
The convenience has a price that rarely appears on the figure. Peptides made this way are normally screened while still tethered, and are not individually purified or individually assayed for identity; the population at a spot includes truncation and deletion sequences whose share grows with peptide length. Where that matters, a cleavable linker allows the material at a spot to be released and analyzed separately. Both points are ordinary bench practice rather than published findings, and are stated here without citation.
Higher density, same physics
Printed formats keep the flexibility while moving to glass. One low-cost protocol synthesizes peptides separately and prints them, interrogating up to 1,536 overlapping peptides derived from the target antigens on a single microslide, with each peptide carrying a polyethylene glycol aminooxy terminus to improve peptide solubility, orientation, and conjugation efficiency to the slide surface [3]. In situ photolithographic synthesis goes further. A high-density peptide array technology with a theoretical upper limit of 2 million different peptides per array of 2 square centimeters was used to lay out more than 200,000 peptide fields in a single array on a microscope slide, holding all possible overlapping 15-mers from human serum albumin together with complete single substitution scans at each position of each 15-mer, probed with a commercially available polyclonal rabbit anti-albumin preparation [4].
Density changes the statistics, not the physics: a million short linear peptides on glass are still short linear peptides.
Linear versus conformational epitopes
A scan presents contiguous sequence. An antibody raised against a folded protein may recognize residues that are neighbors in space but far apart in sequence, and no library of linear peptides contains that surface at all.
The subtler problem is that a peptide can bind well without telling you the conformation in which the site is recognized. For one monoclonal antibody raised against a peptide derived from staphylococcal nuclease, the eliciting peptide and the cognate protein bound with similar affinity — a dissociation constant of 0.30 micromolar for the peptide and 0.42 micromolar for the protein — but with sharply different thermodynamic signatures: peptide binding was characterized by a large enthalpic contribution, minus 7.7 kilocalories per mole, whereas protein binding was dominated by a large entropic contribution, minus 7.2 kilocalories per mole for the entropy term. The X-ray structure of the peptide-Fab complex showed the bound conformation of the peptide differing from the conformation of the corresponding loop region in crystal structures of the free protein [5]. Similar affinity, different route: a scan that scored that peptide as a strong hit would have been right about the binding and silent on the conformation.
Constrained formats try to reach past the linear limit by chemically linking peptides onto scaffolds that hold them in a shape less free than a bare linear chain. A study of infectious bronchitis virus spike protein used arrays of this kind, built from S1 peptides of three circulating strains and screened with sera from chickens immunized with recombinant virus in two independent vaccination and challenge trials; the screening led to the identification of 52 immunogenic epitopes, which the authors assigned into six overlapping epitope binding regions [6]. That is a real extension of the format, not a general solution: the scaffold geometry is a hypothesis about the antigen rather than a measurement of it.
Turning a hit into a footprint
A scan localizes; it does not resolve. Two follow-on libraries do the resolving, and both are standard [1]. A truncation series shortens the hit from each end to find the minimal length still bound. A substitutional analysis replaces each position in turn and asks which replacements are tolerated.
The size of a complete substitution scan is arithmetic worth doing before ordering one. A 15-mer with each of the 20 common amino acids placed at each of the 15 positions is 15 times 20, or 300 sequences, of which 15 reproduce the parent because they restore the native residue at that position, leaving 285 changed sequences — and that is per hit peptide. In the albumin work above, exactly this design was run at each position of every 15-mer in the scan rather than only at the hits, and more than 20 linear epitope candidates were identified and characterized to the level of which amino acids in which positions were needed, or not needed, for antibody interaction [4].
What the intensity is
Spot intensity is a rank, not a dissociation constant. Using one monoclonal antibody as a model system and a set of peptides possessing different affinities for it, SPOT signal intensities were shown to assign spots accurately to three different binding affinity classes, with the dissociation constants of the intermediate region falling between pKdis 5 and pKdis 7; altering the experimental parameters caused a directed change of that region, and accuracy improved when regional trends in the membrane surface were taken into account [7]. Converting those reported bounds, pKdis 5 is a dissociation constant of 10 micromolar and pKdis 7 is 0.1 micromolar, so the intermediate class those authors describe spans two orders of magnitude, with anything tighter than 0.1 micromolar and anything weaker than 10 micromolar falling into one of the two flanking classes rather than into it. The same work noted that applied concentrations of the binding partners do not directly correspond to effective concentrations in the assay [7]. Three classes on one antibody is a modest result, and not a license to read a dissociation constant off a scanner.
The controls that separate signal from surface chemistry
Each control below exists because something other than antigen recognition can light up a spot.
The detection reagent alone. Run the array with secondary reagent and sample buffer but no primary sample. In one study, slides containing 10,000 to 23,000 synthetic peptides were incubated with secondary antibody in sample dilution buffer for this purpose; artifacts attributable to slide architecture were removed before the control peptides were identified, and the authors reported that human immunoglobulin controls, which had responded well on earlier prototype microarrays, did not function consistently on the newer-generation slides [8]. They selected three peptides on each of two array designs, each with a mean response index of at least 5 and a coefficient of variation less than 15%, as replacement controls [8]. The general lesson is that a positive control is valid for the slide chemistry on which it was qualified, not permanently.
Local background, spot by spot. Slides are not uniform. A validation study of peptide epitope microarrays concluded that the ratio of foreground to background signal, on a log scale, provides an appropriate response index, and used replicated assays of samples from five individuals with linear mixed models to investigate the sources of variability, finding high reproducibility of the assay between operators, days and experiments [9].
Peptides from a related protein you do not expect to be recognized. The albumin array carried, as specificity controls, all possible 15-mers from bovine and from rabbit serum albumin alongside the human ones. As the authors expected, moderate cross-reaction was identified with some of the bovine peptides, whereas no cross-reaction was observed with the rabbit peptides [4]. Controls of this kind earn their space because their expected outcome is asymmetric — one set should light up a little, the other not.
Sequences that are not from anything. Near-random peptides bind serum antibodies. In one study, serum from five infectious disease cohorts and an uninfected cohort was profiled against 122,926 peptide sequences selected quasi-randomly to represent an even but sparse sample of the combinatorial sequence space; the authors note that arrays of this kind allow detection and discrimination of humoral immune responses even though their peptides carry near-random sequences that were not designed to mimic biological antigens [10]. The consequence is a design rule rather than a claim those authors make: intensity on a peptide, by itself, does not establish that the sequence is part of an epitope on your target.
The tether. How a peptide is attached changes whether it can be bound at all. One reported method couples biotinylated peptides to NeutrAvidin in solution and microdispenses the peptide-NeutrAvidin complexes onto activated glass; in work covering 54 peptides derived from immunodominant linear epitopes of four protein sources and their cognate monoclonal antibodies, spacer molecules of different type and length were compared for that NeutrAvidin-mediated presentation, and in that format the incorporation of a minimum spacer length was found to be imperative for antibody binding, whereas the peptide immobilization direction had only secondary importance for antibody affinity and binding [11]. That criterion belongs to the presentation format it was measured on, not to peptide immobilization in general. A no-spacer version of the same peptide is worth running once per surface chemistry, not once per experiment.
Surface density. Density is a variable, not a constant of the platform. Using a milk allergen as the model and a chemoselective click immobilization onto a clickable polymeric coating, fine tuning of the surface peptide density was reported to play a crucial role in exploiting oriented and multiple peptide display, and surface plasmon resonance imaging on the most reactive epitope of that series gave the tandem construct an approximately twofold increased binding capacity with respect to the linear and branched forms [12]. Two spots that differ only in how much peptide they carry are not comparable to each other.
What a scan does not settle
Absence is the weakest output the method produces. A region that gives no signal may hold no linear determinant, or may hold one the format failed to present — split by the offset, held in the wrong conformation by the tether, crowded by density, or missing a post-translational modification that the synthetic peptide does not carry. A scan cannot distinguish among these, and a negative should be reported as a negative of that array rather than of the antigen.
Nor does the literature tell you what fraction of antibodies against an arbitrary protein have epitopes a linear scan could recover. A scoping review of peptide microarray applications in human infection research charted 95 articles from 103 studies and found that 77.7% of those 103 studies related to mapping B-cell linear epitopes, with two studies reporting T-cell epitope profiling [13]. That is a description of what the field publishes using the format, not an estimate of how much of an antibody repertoire the format reaches.
Data handling is unsettled in a related way. The validation study quoted above opened by observing that there was then no accepted approach to guide data analysis, with researchers using a wide variety of statistical methods and software tools [9]. Per-slide normalization, an explicit replicate structure and correction for multiple comparisons have since become expected in practice, but they are convention rather than a standard that anyone enforces.
Reporting a scan someone else can read
Nothing in this list is a finding; it is the minimum that makes a scan reproducible, and it is offered as convention. Peptide length and offset, and whether the scan is exhaustive or sampled. Support and coupling chemistry, spacer identity and length, and whether surface density was measured or assumed. Blocking buffer and wash stringency. Probe identity, source and concentration, and the detection reagent. Replicate structure, within slide, between slides and between days. How background was defined and subtracted. Which controls were physically on the array, including the no-primary set and the irrelevant-sequence set. And whether hits were independently re-synthesized and re-tested in a format that does not share the array's surface chemistry, because a hit confirmed only on the surface that produced it has been reproduced, not validated.