You take the segment of a protein that makes the contact, synthesize it, and measure it against the same partner. The protein bound at low nanomolar. The peptide binds at tens of micromolar, if at all. That outcome is common, and the size of the gap is usually explainable by the reasons below. It is not by itself evidence that anything went wrong with the synthesis — but nor does it rule that out, so confirm the peptide's identity, purity and concentration, and check how it has been handled, before settling on a biological explanation.
Affinity ratios sound dramatic until you convert them. At 25 °C, a tenfold change in Kd is worth about 1.36 kcal/mol. So:
| Protein → peptide | Difference in binding free energy at 25 °C |
|---|---|
| 10× weaker | about 1.36 kcal/mol |
| 100× weaker | about 2.73 kcal/mol |
| 1000× weaker | about 4.09 kcal/mol |
These figures describe the overall difference in free energy. They do not identify its molecular causes, and in particular they cannot be converted into a count of lost contacts: what an individual hydrogen bond or side chain contributes to binding depends on its structural surroundings, on desolvation, and on what the rest of the interface is doing. The same 1.36 kcal/mol can represent one interaction, or the small remainder of several that partly cancel.
The value of the conversion is perspective. A hundredfold loss sounds alarming as a ratio and modest as an energy, and the energy is the more useful way to think about how much there might be to recover. It does not tell you what to change.
Inside the folded protein, the recognition sequence is often held close to the conformation the partner requires, stabilised by the structure around it. The same sequence as a free peptide is an ensemble of rapidly interconverting shapes and may spend far less of its time in a binding-compatible one; the conformational freedom it gives up on binding is then subtracted from the binding energy. This is not a universal mechanism, and it has to be established for a particular interaction rather than assumed: some peptides bind while largely disordered and become ordered only in the complex, others bind without ever settling into one rigid structure, and the corresponding region of the intact protein is itself sometimes flexible.
The clearest evidence comes from constrained peptides, where the constraint is deliberately added to pay that cost in advance — described in this literature as “prepaying” the entropic cost of binding, a convenient shorthand rather than a claim that the thermodynamics are settled before the partner arrives. Stapling a helical peptide pre-organizes it, and the practical consequences for affinity, protease resistance and cell entry are the subject of a long literature.1 A study that separated the contributions found that the entropy of the free peptide was the dominant term: a stapled inhibitor of the PSD-95 GK domain bound about 25-fold more tightly than its linear counterpart, improving from tens of micromolar to 1.36 µM, and the gain came mainly from the constrained helix being pre-formed in solution rather than from anything happening in the complex.2
Constraint is not automatically an improvement. The same literature contains peptides that were rigidified into the wrong shape and lost affinity. A constraint helps when it stabilizes the bound conformation; it hurts when it stabilizes something else. The PSD-95 study above makes the point within a single series: two further stapled analogues gained even more entropy than the successful one, yet bound no better, because their looser bound complexes cost enough enthalpy to cancel the gain.2 One study on Cbl(TKB)-binding peptides is titled for exactly this paradox.3
Binding interfaces are usually broader than the stretch that looks most important in a structure. Residues on either side of the chosen segment contribute contacts, and some interfaces draw on residues that are far apart in sequence and brought together only by folding. A linear peptide reproduces none of the latter and only as much of the former as you included.
Trimming can cost a great deal of affinity — but not always, and the direction is not predictable from length alone. The Cbl(TKB) system is a useful caution against assuming otherwise. There, a pentapeptide bound its target roughly fourfold better than the twelve-residue peptide it had been cut down from, which the authors attributed to a smaller desolvation penalty for the shorter sequence. In the same system the large affinity losses, up to two or three orders of magnitude, tracked deviation of the backbone away from the bound poly-proline II geometry rather than the removal of flanking residues.3 So the choice of peptide boundaries deserves as much scrutiny as the synthesis and the sample handling — but what matters is whether the cut removes real contacts or disturbs the conformation, not how many residues came off.
If the parent protein is dimeric, oligomeric, or engages the partner at two sites, its apparent affinity includes an avidity contribution. A monomeric peptide has no access to that. The comparison is then not peptide versus protein but one contact versus several, and avidity can be worth orders of magnitude on its own. This is worth establishing before concluding anything about the sequence.
Several routine, unglamorous problems produce exactly the symptom of weak binding:
These are worth excluding before any conclusion about biology, because each of them is cheap to check and each produces a result that looks like a real affinity difference.
A surface-based measurement with multivalent capture can report an apparent affinity tighter than the true 1:1 value, because rebinding and avidity at the surface both favour the complex. This is a possibility to establish rather than assume: it depends on the binding format, the surface density and the actual stoichiometry. The opposite error also occurs: mass transport limitation and non-specific binding can flatter or spoil a fit. If a peptide looks surprisingly good on a chip, an orthogonal solution measurement is the way to confirm it.
If structure or sequence suggests the interface runs past your boundaries, test a short series with different native flanking lengths rather than one guess — four to six residues each side is a reasonable starting point, but the right length is system-dependent. Compare binding and solubility against the original peptide: extra residues sometimes add nothing, and can change solubility, conformation or accessibility.
Cyclization, a disulfide or a staple can bias the ensemble toward the bound conformation — but check what the constraint actually did before interpreting the affinity. Circular dichroism in your assay buffer is the usual first look and is informative about helicity and gross secondary-structure content; it will not establish the precise three-dimensional binding conformation, which needs NMR or another structural method.
If the parent binds multivalently, a multivalent peptide presentation is the fair comparison. If you want the intrinsic affinity of one site, keep everything monovalent and say so.
Use net peptide content or amino acid analysis to set the concentration rather than gross weight, and check for aggregation before assigning a Kd. Concentration errors propagate directly into affinity.
Weak binding is not always a problem to solve. A micromolar binder remains useful for epitope mapping, competition assays, alanine scanning and specificity profiling, provided the assay has the sensitivity for it and the peptide stays soluble at the concentrations required, because what matters in those experiments is the difference between peptides rather than the absolute affinity of any one of them. The affinity gap only becomes a problem when the peptide itself has to function as an inhibitor or a probe at low concentration.
Please email your project details to sales@lifetein.com or use our online quotation form. We can advise on where to place the boundaries of the peptide, whether a constraint is worth attempting, and which controls make the result interpretable.
Sequences supplied by customers are treated as confidential and are never disclosed or reused.
All LifeTein peptides are supplied for laboratory research use only. They are not for human or veterinary use, and not for diagnostic or therapeutic use.