Automated sample preparation for analytics

Every quantitative analysis begins with a transfer that is not measurement: the analyte has to be moved out of a matrix it is dissolved in and into a form the detector can accept. Almost all of the total variance of a chromatographic or mass-spectrometric result is generated here, before the instrument sees anything. Understanding automated sample preparation means understanding the partition equilibria it is running, not the robot arm executing them.

Extraction is a partition, and it is never complete

Solid-phase extraction works because a molecule distributes between a mobile liquid and a stationary sorbent according to a partition coefficient set by the pair of them. On a reversed-phase C18 sorbent, retention comes from dispersion interactions with the alkyl chain and is defeated by raising organic content; on an ion-exchange sorbent it comes from a charge pairing that is defeated by pH or ionic strength. Selectivity and recovery pull against each other in every one of these chemistries: a wash strong enough to strip interferences also strips some analyte, so the method developer is choosing a point on a trade-off, not eliminating one. Mixed-mode sorbents that combine hydrophobic and ionic retention exist precisely because a single mechanism does not give enough orthogonality against a complex matrix.

Protein precipitation with cold acetonitrile is the crude alternative — it collapses protein solubility by stripping the hydration shell — and it is crude in a specific way: it removes protein and leaves phospholipids, which are the dominant cause of the next problem. Magnetic-bead nucleic acid capture works on a third principle entirely: in high polyethylene-glycol and salt, volume exclusion crowds DNA out of solution onto a carboxylated bead surface, and lowering that crowding releases it. The size cut-off tuned by PEG concentration is a solubility effect, not affinity.

The matrix does not disappear

What survives clean-up still competes at the detector. In electrospray ionisation, co-eluting phospholipids and salts compete for charge and surface area on the evaporating droplet, suppressing analyte ionisation — often by more than half, and variably from sample to sample. This is why an internal standard, ideally a stable-isotope-labelled analogue that co-elutes and suffers identical suppression, is not optional in bioanalysis. Regulatory bioanalytical method validation under ICH M10 requires matrix effect and recovery to be measured in individual lots of matrix rather than in a pooled average, because the variability between lots is the risk, not the mean.

What a liquid handler can and cannot fix

At small volumes the accuracy limit is physical. Air-displacement pipetting relies on a compressible air column, so a volatile solvent that builds vapour pressure inside the tip, or a viscous or low-surface-tension liquid that wets and retains on the tip wall, dispenses the wrong volume in a way calibration with water never reveals. Positive-displacement and acoustic droplet ejection avoid the air column altogether, the latter using a focused ultrasonic pulse to eject nanolitre droplets from an open well with no contact and no tip.

The honest claim for automation is precision, not accuracy. A robot repeats a protocol with a variance an analyst cannot match, which tightens replicate scatter and makes carryover a measurable, controllable quantity. It does not correct a recovery that is systematically low or a wash that loses a metabolite. An automated biased method is a reproducibly biased method, and it fails validation more convincingly than a manual one.