Designing a Dose-Response Study for Peptides In Vitro

A dose-response curve is the fundamental unit of pharmacological evidence. It is also easy to produce badly in ways that look entirely convincing.

Choose the range before the experiment

The most common failure is a range that misses the interesting part of the curve. A curve that plateaus across every point tells you the range was too high; one that never leaves baseline tells you it was too low. Either way the run is wasted.

Start from literature EC50 values for the receptor and system if they exist, and span at least three log units either side. Without a reference point, a wide orienting run — 1 pM to 10 µM in half-log steps — costs one plate and saves several.

Half-log spacing (1, 3.16, 10, 31.6, 100...) gives good curve definition without excessive points. Full-log spacing is acceptable for orienting runs but produces poorly constrained fits.

Serial dilution mechanics

Serial dilution compounds errors: an error in the first step propagates through every subsequent point. Three habits prevent most of it:

• Mix thoroughly at each step before drawing the next. Peptide solutions are not automatically homogeneous after addition.
• Change tips between steps. Carryover on a tip exterior is a real contribution at the low end.
• Prepare fresh from stock for each experiment rather than storing an intermediate dilution series. Dilute peptide solutions are where adsorption losses bite hardest.

For critical work, independent dilutions from stock at each concentration eliminate error propagation entirely. It costs more material and it is worth it for a definitive curve.

Adsorption is the invisible error

Peptides stick to surfaces. At micromolar concentrations this is negligible. At low nanomolar and below, adsorption to polypropylene tubes, pipette tips and plate wells can remove a substantial fraction of your peptide from solution.

The consequence is systematic, not random: your low concentrations are lower than you think, which shifts the whole curve right and inflates your apparent EC50.

Mitigations:

• Low-binding plasticware throughout the dilution series and the assay
• Carrier protein — 0.1% BSA in the diluent, if your assay tolerates it — occupies binding sites competitively
• Minimise transfers and surface contact time
• Glass for stock storage where practical, silanised if available

Vehicle controls

Every dose-response needs a vehicle control at the highest vehicle concentration present in any well.

If your peptide stock is in DMSO and the top concentration puts 0.5% DMSO in the well, the vehicle control is 0.5% DMSO. Not 0.1%, not "medium only". A 0.5% DMSO effect masquerading as a top-concentration peptide effect is a classic and entirely avoidable artefact.

The same applies to bacteriostatic water: benzyl alcohol at the concentration reached in your top well belongs in the control.

Concentration accuracy

Your x-axis is only as good as your stock concentration, and two corrections are routinely skipped:

Net peptide content. A 10 mg vial of a TFA salt may be 78% peptide by mass. Using the label mass rather than the COA's net peptide figure overstates your concentration by nearly a quarter across the entire curve.

Water content. Lyophilized peptides carry 1–5% residual moisture, and more if handled carelessly. Karl Fischer data on the COA lets you correct.

Neither shifts the shape of the curve, but both shift the EC50 — which is usually the number you are reporting.

Replicates

Technical replicates (same plate, same preparation) measure assay precision. Three per concentration is standard.

Biological replicates (independent experiments on different days, ideally different cell passages) measure whether the result is real. Three independent experiments is the usual minimum for publication.

These are not interchangeable, and treating technical replicates as n=3 for statistics is one of the more common errors in the literature. Technical replicates tell you about your pipetting. Biological replicates tell you about biology.

Fitting the curve

The standard model is the four-parameter logistic:

Response = Bottom + (Top − Bottom) / (1 + 10^((logEC50 − log[A]) × HillSlope))

Four parameters: bottom plateau, top plateau, logEC50, and Hill slope.

Practical guidance:

• Fit log-transformed concentrations, not linear. Fitting linear concentration gives disproportionate weight to the top of the range and produces poor EC50 estimates.
• Constrain the bottom to your vehicle control value where the data justifies it. An unconstrained bottom floating below the control is fitting noise.
• Report confidence intervals on EC50, not just the point estimate. An EC50 of 4 nM with a 95% CI of 1–16 nM is a very different claim from the same value with a CI of 3.5–4.6 nM.
• Look at the Hill slope. A slope far from 1 suggests cooperativity, multiple binding sites, or — more often — a problem with the assay.
• Do not extrapolate. If the curve has not plateaued, the top parameter is unconstrained and the EC50 is unreliable. Extend the range and rerun.

Reporting

A dose-response result that others can evaluate includes:

• The full concentration range and spacing
• The vehicle and its final concentration
• Peptide source, batch number and stated purity
• Whether concentrations were corrected for net peptide content
• n for both technical and biological replicates
• The fitting model, any constraints applied, and confidence intervals
• The reference agonist run in the same system

That last one carries more weight than any other. Potency is comparative. An EC50 without a reference agonist run alongside it in the same system on the same day is a number without a scale.

All materials referenced are for laboratory research use only.