# Downstream purification (bioseparation)

How a protein's size, charge, hydrophobicity and affinity map onto unit operations, why the train runs capture–intermediate–polish, and how the selectivity-versus-throughput tradeoff sets the cost of goods.

Purification is a chain of compromises: every step that sees the product more precisely holds less of it, and the train's economics are decided at its dirtiest end.

Source: https://en.bioecon.ru/docs/bioproduction-equipment/downstream-purification/downstream-purification-bioseparation/
Updated: 2026-09-07



A bioreactor delivers the product as a few grams per litre in a murky aqueous soup of host proteins, DNA, salts and viruses; the specification demands a solid or solution that is essentially one molecule. Nothing between those two states creates material — purification only spends selectivity, and every operation spends it at a different exchange rate against throughput. The design question is never "which method works" — almost all of them work — but which compromise chain loses the least money for this particular molecule.

## The target is a set of coordinates

A protein differs from its impurities along a handful of physical axes, and each axis is a unit operation. Size: size-exclusion chromatography resolves by hydrodynamic volume, membrane cutoffs split by molecular weight. Charge: ion exchange binds the protein when the buffer pH puts its net charge opposite to the resin, and elutes it by shifting pH or salt. Surface hydrophobicity: hydrophobic-interaction chromatography uses high salt to push the protein onto hydrophobic ligands, then lowers salt to release it. Biospecific affinity: a ligand that recognises the molecule directly — an antibody-binding protein, a substrate analogue, a tag — is the only operation that "knows" the target rather than one of its properties. Density and phase: centrifugation and filtration (see ../disc-stack-centrifuges/) handle cells and debris before chromatography earns its keep. The impurity spectrum picks the axes that matter: host-cell proteins differ from the product in all coordinates at once, aggregates differ only in size, charge variants only in charge — the closer the impurity, the fewer the usable axes and the more expensive the step.

## Capture, intermediate, polish

The train is ordered by volume, descending. Capture works on the whole broth: its job is volume reduction and getting the product away from the proteases still in the broth, so it maximises throughput and accepts modest resolution. Intermediate steps remove the bulk of host protein and DNA. Polish removes what is left — aggregates, fragments, leached ligand, close structural relatives — where only high resolution works and throughput hardly matters. Each stage has its own place on the selectivity-capacity-speed surface, which is why the same logic recurs from antibody plants to enzyme plants even when the hardware differs.

## Why selectivity sets the cost

Two arithmetic facts do the economics. First, steps multiply: five steps at 90% yield each deliver barely 60% of the product — yield lost to the number of steps, not to any one bad step — so the fewest, most selective operations win. Second, dilution propagates: a low-selectivity step that lets impurities through forces another step, another buffer exchange, another hold volume, and equipment scales with volume. This is the economics behind affinity capture: an expensive resin that binds one product class near-selectively deletes two or three cheaper steps and all their volumes. Platform processes exist because molecules sharing coordinates share trains — a new antibody reuses an old antibody's purification, and only the polish steps care which antibody it is. And when the product is a small, robust molecule, the whole chromatographic logic can be left behind for distillation, extraction or crystallisation (see ../crystallization-equipment-pharma-api/) — selectivity bought from phase equilibrium instead of from resin.