A result that cannot be repeated is not a finding. This page covers the design decisions that separate the two: what your controls have to rule out, what your n actually counts, and which details another laboratory needs in order to reproduce what you did.
A control is not a formality. Each one exists to eliminate one specific alternative explanation for your result.
| Control | What it is | Alternative explanation it eliminates |
|---|---|---|
| Vehicle | Solvent alone at the same final concentration | "The solvent caused it" |
| Untreated | No addition at all | "Handling and medium change caused it" |
| Positive | A compound with a known effect in this assay | "The assay was not working that day" |
| Negative / scrambled | Inactive analogue or scrambled sequence | "Any peptide would do this" |
| Blank | No cells, reagents only | "The compound interferes with the readout itself" |
The blank deserves particular attention with coloured or fluorescent compounds: a molecule that absorbs at the detection wavelength can produce a dose-dependent signal with no biology behind it whatsoever.
Three wells from one plate seeded from one flask are technical replicates. They estimate how precise your measurement is - pipetting, plate position, reader variation. They are worth running, and they tell you nothing about whether the effect is real.
Averaging technical replicates gives you one better-estimated value. It does not give you three observations.
A separate passage, a separate donor, a separate animal, an experiment run on a different day from a different thaw - these are biological replicates. They estimate the variability that actually matters, because they include everything that differs between genuine repetitions.
Statistics belong at this level. Computing a p-value across technical replicates inflates n and produces confidence the data does not support. It is among the most common reasons a published in-vitro effect fails to replicate.
Plate position is not neutral: edge wells evaporate faster and run at different temperature, which is enough to produce a gradient that mimics a dose response. Randomising treatment across plate positions removes that confound; leaving a tidy left-to-right dose layout builds it in.
Where a readout involves judgement - scoring morphology, counting - the scorer should not know the treatment. Blinding costs a relabelling step and removes an entire class of unconscious bias.
Use logarithmic spacing across at least three to four decades. A narrow linear range routinely misses the transition entirely and returns a flat line.
A curve fit is only constrained if the data reaches a floor and a ceiling. Without both, the fitted EC50 is an extrapolation.
It describes the compound in that assay, cell line and condition set. Comparing EC50 values across papers with different systems compares the systems as much as the compounds.
⚠️ Nominal concentration is not delivered concentration. Adsorption to plasticware, purity below 100%, and net peptide content below 100% all mean the molar concentration reaching the cells is lower than the number written on the tube. See assay methods for why net content governs this and handling for adsorption.
Reproducibility is a property of the write-up as much as of the work. These are the details whose absence most often makes a method unrepeatable.
| Report | Why it changes the result |
|---|---|
| Compound source and lot number | Purity, net content and impurity profile vary lot to lot |
| Documented purity and net peptide content | Governs the true molar concentration |
| Solvent and final vehicle concentration | Solvent alone has biological effects |
| Reconstitution date and storage since | Solution age determines degradation state |
| Freeze-thaw cycle count | Damage accumulates per cycle |
| Cell line, passage number, medium and serum lot | Phenotype drifts with passage; serum is undefined and variable |
| n, and whether it is technical or biological | Determines whether the statistics are valid |
| Exact statistical test and correction applied | Multiple comparisons without correction inflate significance |
Every lot we ship is published with its Certificate of Analysis against its lot number, so the first two rows of this table can be cited directly from the COA library rather than requested.
A dish has no absorption, distribution, metabolism or excretion. Concentration at the cell is set by you, not by an organism.
Monoculture lacks the architecture, matrix and cell-cell signalling that shape a response in tissue.
Immunogenicity and clearance simply cannot be observed in most in-vitro systems.
An in-vitro finding is a hypothesis about a biological system, not evidence about an organism. Stating that limit explicitly is what keeps the result usable to the next reader.
Research Use Only. This page is methodological reference material for qualified laboratory professionals working in vitro. It is not a protocol, not medical or veterinary guidance, and contains no dosing recommendation. Products supplied by On Top Labs are not drugs, supplements or medical devices.