There’s a pattern that keeps happening. Upstream works. Titers look great. And then someone asks: “What’s the purification strategy?”
It’s three months from the IND filing. The molecule is locked. The separation train has to work around it now — not with it.
It doesn’t have to go this way.
The Cost of Designing Late
Separation problems are cheaper to solve before they’re locked in. Your expression system, fermentation conditions, and harvest method all constrain what’s downstream of them. A CHO cell line at high cell density creates a completely different clarification problem than a microbial fermentation — and those differences cascade through every unit operation that follows.
Design downstream late and you’re paying for ad hoc experiments that a proper flowsheet simulation would have caught at the whiteboard stage.
The data you need to sketch a first-pass separation train — feed composition, target molecule properties, rough titer — is usually available in week one.
What “Early Process Design” Actually Means
It’s not about locking in vendors or buying equipment. It’s three things:
Characterize your feed stream
What are the major impurities? Cells, host cell proteins, DNA, small molecules? Order-of-magnitude concentration estimates are enough to start eliminating options.
Map your target’s property space
Molecular weight. Isoelectric point. Solubility. Charge at process pH. These define which unit operations give you selectivity — and which ones just concentrate everything together.
Run the separation logic on paper
Given a feed with component X at concentration Y, what does a 2-step or 3-step train look like? What yield and purity are theoretically possible? Where are the failure modes?
The Unit Operation Order Matters
For most biotech molecules, the separation train follows a logic:
Cells and debris first
Centrifugation or depth filtration. No downstream unit operation handles whole cells gracefully.
Gross size fractionation
Microfiltration (0.2–0.45 µm) for particles, ultrafiltration (1–1000 kDa) for macromolecules. Match the MWCO to your target, not the impurity.
Charge-based capture
Ion exchange when your target has a different net charge from the main impurities at a given pH. That requires knowing your pI and the impurities’ pI.
High-resolution polishing
Affinity, size exclusion, or reverse phase as a final step — after everything else has done the heavy lifting.
Concentration and formulation
UF/DF before fill. Don’t try to crystallize or dry a dilute stream.
Running ion exchange before clarification fouls columns with particulates. Running size exclusion before concentration dilutes an already dilute stream. The order isn’t arbitrary.
Run the Mass Balance First
Before equipment sizing or cycle time: run the mass balance.
The product of yield fractions at each step is your total recovery. That math is unforgiving at scale. Losing 30% of a $10,000/g product in a single step is a significant cost driver.
| Per-step yield | 2 steps | 3 steps | 4 steps |
|---|---|---|---|
| 90% | 81% | 73% | 66% |
| 80% | 64% | 51% | 41% |
| 70% | 49% | 34% | 24% |
Selectivity — the ratio of product enrichment to impurity enrichment — tells you whether a step is doing useful work. A step with selectivity near 1.0 concentrates impurities as fast as it concentrates product. That step needs to be replaced or repositioned.
Simulation Is a Design Tool, Not a Documentation Tool
Most teams treat simulation as something you do after design is complete — to generate regulatory documentation. Wrong end of the pipe.
The value is in the what-if phase. What happens to purity if the ion exchange column loads 20% over design? What if upstream media changes shift the feed composition? What’s the pH at the outlet of the mixing step — and will that precipitate your protein?
untangle.bio was built for this phase. Drag-and-drop unit operations. Real mass-balance calculations with pH tracking and precipitation checks. Instant what-if analysis. Sketch a 4-step train in minutes and see yield and purity at each step — before ordering a single column.
The Loop That Works
Rough flowsheet → mass balance → identify the weak step → redesign it → repeat. Short iterative loops, starting in week one.
One long loop that only closes when you’re already in the pilot plant and constraints are locked.
Start earlier. Use the data you already have.
Design your separation train before the constraints lock in
Drag-and-drop unit operations, real mass-balance calculations, pH tracking, and instant what-if analysis — all in the browser.
Try it free →