How to Choose the Right Separation Technology for Your Molecule

You’ve got a molecule that works. Now someone wants a purification strategy.

If you’ve spent your career on the biology side, this question can feel like it belongs to someone else. But the separation strategy you pick at the bench stage shapes everything downstream: yield, purity, cost, scale-up risk. Getting it wrong early is expensive later.

Here’s a practical framework. No process engineering degree required.

Start With Your Molecule’s Properties

Every separation technology exploits a physical difference between your target and the things you want to remove. The first question isn’t “which technology?” — it’s “what’s different about my molecule?”

Property	Why it matters	Separation technology it informs
Molecular weight	Defines size relative to membrane cutoffs and SEC range	Ultrafiltration, nanofiltration, size exclusion chromatography
Isoelectric point (pI)	Charge at any given pH — drives binding to ion exchange resins	Ion exchange (CEX, AEX), precipitation design
log P (hydrophobicity)	Partitioning between aqueous and hydrophobic phases	Reverse phase, HIC, liquid-liquid extraction
Solubility	Sets crystallization feasibility; drops near pI for proteins	Crystallization, precipitation
Charge at process pH	What actually matters — depends on buffer, not just pI	Ion exchange binding conditions, precipitation risk

For proteins, the isoelectric point is the most useful single number. It tells you what pH to run ion exchange at, whether you’ll have aggregation problems during low-pH viral inactivation, and where not to operate if you want to avoid precipitation.

The Four Main Separation Categories

Ultrafiltration

Size-Based · High Throughput

UF membranes separate by molecular weight cutoff (MWCO). The rule: your target should be well above or well below the cutoff — never near it.

Retaining your target: use a MWCO at 10–30% of target MW
Letting your target pass: use a MWCO at 3–10× target MW
Washing out small-molecule impurities: diafiltration — 10 diavolumes removes ~99.995% of a permeable impurity

UF is excellent for concentration and buffer exchange. Not high-resolution — if your impurity has a similar MW to your target, a membrane won’t distinguish them.

Ion Exchange Chromatography

Charge-Based · High Resolution

Mode	Exploits	Best for	Avoid when
AEX (anion exchange)	Positively charged resin; negatively charged molecules bind	Acidic proteins, DNA removal, endotoxin clearance	Target has net positive charge at process pH
CEX (cation exchange)	Negatively charged resin; positively charged molecules bind	Basic proteins (mAbs, lysozyme), charge variants	Target has net negative charge at process pH
SEC (size exclusion)	Pore size — hydrodynamic radius	Polishing, aggregate removal, MW confirmation	Primary capture step or high-throughput scale
Reverse phase	Hydrophobicity — partition to non-polar resin	Small molecules, peptides, antibody-drug conjugates	Folded proteins (causes denaturation in most cases)

Affinity Chromatography

Specific Binding · Single-Step Purity

Affinity uses an immobilized ligand that specifically binds your target. Single-step purities above 95% are routine. Protein A for antibodies, IMAC for His-tagged proteins. Expensive resin — reserved for high-value molecules where the economics justify it.

Crystallization

Solubility-Based · Underused for Small Molecules

Crystallization exploits solubility differences. Most useful when: the target has low solubility (organic acids, amino acids, many small molecules), you need a dry solid final form, or you want to skip column-based purification entirely. Requirement: reach supersaturation without co-precipitating impurities.

Centrifugation & Filtration

Particle Removal · Clarification First

For any fermentation or cell culture process, remove cells before anything else. Options: disc stack centrifuge (high throughput, continuous), depth filtration (size exclusion plus adsorptive capture), microfiltration 0.2 μm (clean clarification, fouls quickly at high cell density).

A Decision Framework

Working out your first-pass strategy? Answer these questions in order.

1

What is the feed source?

Fermentation or cell culture → clarification first (centrifugation or microfiltration). Synthetic or enzymatic reaction → may go directly to a capture step.

Clarification before membrane or chromatography

2

What is the target molecular weight?

<1 kDa → consider NF, crystallization, or reverse phase. 1–50 kDa → UF capture followed by IEX polishing. >50 kDa → standard bioprocess train (clarify → affinity or IEX → UF).

MW is the primary branching point

3

What is the main impurity?

Similar size to target → charge- or hydrophobicity-based separation needed. Very different size → a membrane step does the heavy lifting.

Defines which physical property to exploit

4

What purity is required?

>95% → at least two chromatography steps. >99% (pharmaceutical) → three steps plus viral clearance. Food/industrial grades tolerate simpler trains.

Purity target sets the number of steps

5

What is the intended scale?

Bench scale: almost anything works. Pilot or commercial: avoid SEC as the primary capture step (low throughput, dilutes your stream). Prefer bind-and-elute formats for scale-up.

Scale eliminates certain options early

The pH Question Scientists Miss

pH affects more than enzyme activity. For charged molecules, it controls:

Net charge — which determines ion exchange binding and whether your molecule sticks to the resin at all
Proximity to pI — precipitation risk for proteins (minimum solubility at the isoelectric point)
Dissociation state — solubility for weak acids and bases follows Henderson–Hasselbalch

          Organic acids and pKa: Lactic acid, citric acid, and gluconic acid have pKa values between 3–5. Above their pKa they are dissociated (A−) and highly soluble. Below it they are protonated (HA) and potentially near their intrinsic solubility limit. Run a separation at the wrong pH and your product crashes out of solution before you’ve purified anything.
        

Check the pH at every step against the pKa or pI of every component that matters. Before you run the experiment.

A low-pH viral inactivation step can be excellent for virus clearance and catastrophic if your protein’s pI is 4.5. Each step affects what comes next — don’t design them in isolation.

Don’t Design Steps in Isolation

The interaction between steps is where processes fail. A classic failure mode: optimise each unit operation independently, assemble them into a train, and discover that the eluate from step 2 is at the wrong pH for step 3 to bind at all.

untangle.bio lets you wire up unit operations into a flowsheet, run the mass balance across the whole train, and see pH propagation, precipitation warnings, and per-step purity and yield — before you go near the bench.

That systems-level view is worth more than any single technique optimisation.