ICH Q2(R2) separates precision into three tiers, and the separation is instructive for work that has nothing to do with regulatory submission. Repeatability is what one analyst achieves on one instrument across a short interval. Intermediate precision adds variation across days, analysts and equipment inside one laboratory. Reproducibility covers agreement between laboratories. The three numbers are rarely equal, and the gap between the first and the third is where most unexplained inter-laboratory disagreement lives.
A serial dilution is a precision problem of exactly that shape, performed by hand, usually without any of the three being characterised. It is also the step that converts a weighed quantity of solid into every concentration the experiment will report, which makes it the point at which a systematic error becomes invisible by becoming universal.
The arithmetic that makes serial dilution unforgiving
The distinction worth understanding is between serial and independent dilution, because the two accumulate error in fundamentally different ways.
In an independent dilution scheme, each working concentration is prepared separately from the stock. An error in one preparation affects that concentration and no other. The errors are independent, they do not accumulate, and a single bad transfer produces a single outlying point that is usually visible as such.
In a serial scheme, each step is prepared from the one before it. An error early in the series propagates through everything downstream, and it propagates multiplicatively rather than additively. Three transfers each accurate to within two per cent do not yield a final concentration accurate to within two per cent; the uncertainties compound, and by the fifth or sixth step the accumulated uncertainty on the lowest concentration is substantially larger than on the first.
Worse, a systematic error, as opposed to random scatter, does not announce itself at all. A pipette delivering consistently low across the whole series produces a set of concentrations that are all wrong in the same direction by a progressively larger factor. The resulting curve is smooth, internally consistent and shifted. Nothing in the data indicates a problem, because the problem is the same at every point.
This is why a laboratory that cannot reproduce another laboratory’s half-maximal value, using the same compound and the same assay, is often comparing two correctly-executed experiments performed on two differently-calibrated sets of pipettes.
Where the error actually enters
Four contributors dominate, and they are not equally weighted.
The initial weighing is the largest in proportional terms, because the quantities involved are small. Weighing a few milligrams of a hygroscopic lyophilised solid on a balance in an ordinary laboratory environment carries uncertainty that is a meaningful fraction of the quantity, and that uncertainty sits at the head of the series where it propagates furthest. Material picking up atmospheric moisture while the balance door is open has changed before it reaches the solvent.
Pipette calibration is the one most often assumed rather than verified. Air-displacement pipettes drift, and they drift fastest at the bottom of their nominal volume range, which is precisely where serial dilution work tends to operate. A pipette set to deliver a small volume near the lower limit of a larger-capacity instrument is the single most common source of silent systematic error in dilution work.
Mixing between steps is the contributor most often rushed. A transfer into an unmixed or incompletely mixed intermediate draws from a concentration gradient rather than from a homogeneous solution, and the error that results is neither random nor predictable.
Adsorptive loss to labware is the peptide-specific contributor and it is concentration-dependent, which makes it the most awkward of the four. Loss is proportionally greatest at the lowest concentrations, which means it does not shift the whole curve uniformly but distorts its shape, flattening the low end in a way that mimics a genuine biological feature.
Why the curve shape matters more than the absolute values
Many experiments are internally comparative. A laboratory comparing two compounds, or one compound across two conditions, cares about relative behaviour rather than absolute concentration, and a uniform systematic error cancels out of that comparison.
That tolerance disappears the moment a number leaves the laboratory. A half-maximal value quoted in a paper is an absolute claim, and it inherits every systematic error in the dilution chain. Two laboratories reporting values differing by a factor of two or three, for the same compound in the same assay format, are not necessarily disagreeing about biology.
The tolerance also disappears when the error is concentration-dependent rather than uniform. Adsorptive loss does not cancel in a comparison, because it affects the two compounds differently according to how adsorptive each is.
What a defensible dilution scheme looks like
Several practices reduce the accumulated uncertainty, and none requires new equipment.
Prepare independently rather than serially where the number of concentrations makes it practical. The additional stock consumption buys independence of errors, and for a short series it is usually the better trade.
Where a serial scheme is necessary, keep the number of steps low by using larger dilution factors per step rather than many small ones, and avoid operating any pipette near the bottom of its volume range.
Verify gravimetrically. Weighing a transfer, rather than trusting the volume setting, converts a volumetric assumption into a measurement, and it is the only practical way to detect a drifting pipette at the bench.
Mix deliberately at every step, with a defined procedure rather than an improvised one, and record what the procedure was.
Use low-binding consumables where the series extends to low concentrations, and prepare the low end fresh rather than storing it, because adsorptive loss continues during storage.
And verify the stock independently where the result is load-bearing. An absorbance reading on the stock, against the compound’s known extinction behaviour, is a quick check that the weighing and the first dissolution produced what was intended.
How it works in practice
Every one of those practices starts from a stated quantity of material, which is where supplier documentation enters.
A stock concentration calculated from the vial label is only as good as what the label means. A quantity stated as total fill weight, for material that is a salt carrying residual water, overstates the peptide present. A quantity stated as peptide content does not. The difference between the two is not a rounding error for a trifluoroacetate salt, and it sits at the head of the dilution series where it propagates through every concentration reported.
For a research compound such as Bluum Peptides eloralintide, supplied for laboratory research use, the documentation checks that bear directly on this are therefore specific: is the stated quantity peptide content or fill weight, is counterion identity and content disclosed, and is water content reported for the lot in hand. For laboratory research use only. Not for human or veterinary use. Where all three are present, a laboratory can calculate a stock concentration that means what it says. Where they are absent, the first number in the chain is an estimate, and no amount of care downstream recovers it.
Supplier solubility guidance matters for the same reason, because it determines the achievable stock concentration and therefore how many dilution steps the series needs.
What the record should contain
The information that makes a dilution reproducible is known at the bench and rarely survives into a write-up.
A method section that supports replication states the lot number, the quantity weighed and on what basis, the solvent and the stock concentration, the dilution scheme as serial or independent with the factor per step, the consumables used where binding matters, and the interval between preparation and use. That is six lines. A typical description gives the supplier, the catalogue number and the final concentrations, which identifies a product line and the intended endpoint while omitting everything that determined whether the endpoint was reached.
Where this is heading
Automation has removed some of this problem where it is available, since liquid handlers are more repeatable than hands and keep their own records. It has not removed the weighing uncertainty at the head of the chain, nor the adsorptive loss at the tail.
Reporting conventions have moved towards lot-level reagent identification, and the natural extension is preparation-level description. Some journals now ask for solvent proportion in assay descriptions explicitly. Dilution scheme is not yet a standard field, and in its absence the discipline sits with the laboratory.
The argument for adopting it regardless is the gap between repeatability and reproducibility that ICH names. A laboratory that characterises its own dilution practice knows which of its disagreements with other laboratories are worth investigating.
Conclusion
A serial dilution propagates error multiplicatively from its first step, and a systematic error propagates invisibly because it affects every point in the same direction. The weighing at the head of the chain carries the largest proportional uncertainty, pipette calibration carries the most common silent one, and adsorptive loss at the tail distorts curve shape rather than shifting it.
The defences are independent preparation where practical, fewer and larger steps where it is not, gravimetric verification instead of volumetric trust, and a label that states peptide content rather than fill weight so the first number is right. The last of those is the supplier’s to provide, and it is worth checking before the order rather than after a curve refuses to match somebody else’s.


