Figure 1. Schematic representation of the a) batch and b) rapid cycling parallel multi-column chromatography (RC-PMCC) process. Non-loading steps: wash, elution, regeneration, and equilibration are performed one at a time. (Adapted from Schmitz et al., 2024)
1. Define the Architectures Being Compared
Here, batch chromatography means a conventional cyclic packed-column process: one column performs loading, washing, elution, and regeneration in sequence, with product collected as discrete pools. It does not mean mixing loose resin with sample in a vessel. That distinction matters because batch adsorption and cyclic column operation have different transport and scheduling behavior.
Multi-column chromatography is a family of arrangements using two or more columns with coordinated flow paths and phase timing. Some configurations alternate independent columns, while others connect columns during loading so a downstream bed captures material leaving an upstream bed. Different architectures provide different benefits and should not be treated as one universal technology.
Distinguish equipment mode from process continuity
A multi-column capture step may accept feed continuously while releasing product in intermittent eluate pools. It does not automatically make the entire downstream process continuous. Define how feed enters, how product exits, and how other unit operations connect before using continuous operation as a claimed benefit.
2. Understand Where Additional Resin Utilization Comes From
In single-column capture, loading often stops before substantial target breakthrough to protect recovery. At that moment, part of the bed may remain underutilized because the mass-transfer front has not fully traversed the column. The gap between conservative dynamic loading and more complete utilization creates a potential opportunity for interconnected columns.
During a lead-and-lag loading sequence, a downstream column can capture target escaping from the lead column. The lead bed can then be loaded more fully before switching roles or entering recovery phases. This principle can improve utilization, but the gain depends on breakthrough shape, residence time, feed composition, and the acceptable leakage strategy.
Do not equate higher loading with higher accepted output. Impurity displacement, elution quality, irreversible retention, or product instability may limit useful loading before target capacity is exhausted. Confirm the complete separation and mass balance under the intended architecture, including what happens to product that moves between beds.
3. Choose Applications with a Plausible Advantage
Multi-column capture can be attractive when feed supply is sustained, resin is costly, conventional loading leaves substantial capacity unused, or overlapping phases can reduce equipment occupancy. Relatively predictable feed and robust affinity capture can simplify development, although suitability still depends on the actual target, resin, and process requirements.
Some polishing separations can also benefit from multiple columns, but the separation problem may be different. Resolving closely related variants or controlling impurity fronts can require more elaborate fraction routing and cycle design than capturing a target from a dilute stream. An architecture successful for affinity capture should not be assumed to transfer directly to every polishing step.
For variable research batches, frequent product changes, small annual demand, or uncertain feed availability, a conventional cyclic process may provide sufficient output with simpler implementation. Start with the production problem that needs solving, then ask whether additional columns address it better than a larger bed, shorter phases, or improved upstream scheduling.
4. Balance Loading against Recovery and Regeneration
Develop a phase schedule showing when each column loads, washes, elutes, regenerates, and reequilibrates. The number of columns required depends on the loading interval and the time needed for all other phases, as well as the selected architecture. Adding columns without checking this balance can create idle capacity rather than improved throughput.
Account for buffer delivery, shared pumps, valve availability, collection vessels, and pressure limits. A schedule that is mathematically feasible may be impossible on the selected skid if simultaneous phases require the same flow path or buffer source. Include cleaning contact time and flushing endpoints instead of treating them as freely compressible overhead.
Include realistic variability
Feed concentration or flow changes can shift breakthrough and disturb synchronization. Evaluate how the system responds: change switching time, adjust flow, use a surge vessel, or pause under a defined procedure. The control strategy must maintain product quality across expected disturbances rather than relying only on an ideal periodic cycle.
5. Compare Both Processes on the Same Basis
Use accepted target mass divided by total installed resin volume and elapsed system time when reporting resin-normalized productivity. Include every participating bed, including beds in recovery or regeneration. Comparing output against the volume of only the currently loading column overstates multi-column performance.
Measure campaign output as well as steady-operation productivity. Startup, initial conditioning, transition pools, shutdown, and residual product recovery can affect short campaigns substantially. Define which fractions meet quality criteria and how material outside those pools is handled. Steady-state performance is useful but incomplete when routine operation spends significant time in transitions.
| Metric | Common comparison basis | Avoid this distortion |
|---|---|---|
| Resin productivity | Accepted mass / total resin / elapsed time | Counting only the loading bed |
| Recovery | Accepted mass / total applied target | Omitting startup or shutdown material |
| Buffer use | Total prepared and consumed liquid / accepted mass | Comparing one phase with a complete cycle |
| Capacity demand | Output over the same campaign duration | Using ideal steady operation for only one option |
Apply the same analytical methods and quality limits to both options. A higher output rate achieved by looser pooling or a different impurity endpoint is not an architecture benefit.
6. Design Switching and Monitoring around Product Behavior
Switching may use elapsed time, processed volume, detector signals, or a combination supported by development data. A fixed schedule is simpler but can be vulnerable to feed changes. Signal-based control can adapt, yet requires a measurement that reliably indicates the event of interest under the actual feed matrix.
Absorbance alone may not distinguish target breakthrough from other proteins or background components. Validate the relationship between the online signal and target behavior using suitable offline assays or additional measurements. Define sensor range, response delay, baseline handling, and actions for missing or implausible signals.
Map valve transitions and line hold-up carefully. Routing changes can cause cross-contamination, brief pressure disturbances, or unintended product loss if timing is wrong. Test interlocks, collection logic, and the response to a stopped pump or unavailable column. A multi-column method includes its control sequence, not only its buffer recipes and loading conditions.
7. Confirm Product Quality across Columns and Time
Evaluate product from individual columns and successive cycles before relying on a combined campaign pool. Differences in packing, resin age, or fouling can create unequal behavior even when nominal columns are identical. A composite sample can conceal an isolated excursion or a progressive trend.
Track critical impurities, product integrity, accepted recovery, pool volume, and relevant carryover. Establish how material is identified and traced through interconnected flow paths and pooled collections. The material-accounting approach should support investigation and segregation when an event affects only part of the operation.
Assess cleaning and storage for each bed and shared path. More valves and lines can introduce additional locations requiring flushing and control. Define maximum processing and hold durations with product stability in mind. Regulatory or quality-system obligations depend on the intended application, but sound traceability and a justified response to excursions are practical needs for any architecture.
8. Include Implementation and the Surrounding Process
Compare resin inventory, achieved lifetime, buffer use, cycle timing, hardware, automation, development, training, and maintenance. Improved resin utilization can reduce one cost component while additional control equipment increases another. Use the same annual demand and campaign assumptions, and separate one-time implementation expense from recurring operation.
Consider upstream supply and downstream acceptance. Continuous feed may require a controlled surge strategy, while intermittent eluate may still need collection vessels. Downstream concentration, viral inactivation where applicable, polishing, or filtration can limit overall throughput. A faster capture system creates little benefit if the next operation becomes a severe bottleneck.
Use the resin cost framework to test sensitivity to resin lifetime, campaign length, and accepted recovery. Do not assume that more complete loading extends resin life; exposure, fouling, and cleaning frequency can change in either direction. Base lifecycle economics on representative data or explicit uncertainty scenarios.
9. Use a Staged Development Program
Begin with a well-characterized single-column baseline using representative feed. Measure breakthrough shape, usable capacity, recovery, impurity behavior, and phase durations. These data indicate whether underutilized capacity or nonloading time offers enough opportunity to justify architectural development.
Build a small multi-column demonstration with measured hold-up, reliable switching, and complete material balance. Investigate startup, periodic operation, disturbances, and shutdown. Challenge meaningful feed variation and unequal column behavior where relevant. Retain enough fraction resolution to identify where quality or recovery differs from the baseline.
Before scaling, define the variables that must be preserved and those controlled by the hardware. Column dimensions, residence time, interconnection volumes, detector delay, and switching logic interact. Apply scale-down model principles to establish what the development system can represent, and retain the control configuration as part of the transferable method.
10. Evaluate Startup, Shutdown, and an Unavailable Column
Startup establishes the loading state of each bed and the distribution of material through the connected system. Early cycles may not represent the later repeating pattern. Define when product becomes eligible for the routine pool and how transitional fractions are measured, retained, or rejected. Include their mass in campaign recovery whether or not they are pooled.
Shutdown must address product still present in feed lines, partly loaded columns, and intermediate collections. A recovery sequence may require additional buffer, time, and analytical testing. Compare its value against the quantity and quality of recoverable material rather than assuming that shutdown is identical to a normal cycle boundary.
Plan for a column that cannot return to service
If one bed fails a pressure or packing check, determine whether the architecture can continue at a reduced feed rate, divert feed to a controlled hold, or stop safely. The answer depends on the available columns, phase timing, product stability, and hardware. Do not assume that any multi-column system can simply remove one column without changing the validated sequence.
Use a defined disturbance test or justified simulation to evaluate the proposed response before routine operation. Track where material is located at the event and which subsequent pools could be affected. Demonstrate that the control logic and collection records support the intended segregation and investigation.
These transition scenarios are especially important for short campaigns and valuable feeds. A system with strong steady-operation productivity can lose its advantage if setup and recovery consume much of the available production window. Report both repeating-cycle performance and the complete campaign outcome, and base the architecture decision on the operating pattern the facility will actually use.
11. Make the Architecture Decision Explicit
Choose conventional cyclic chromatography when it meets output and quality requirements with acceptable resource use and provides a practical operational fit. Consider multi-column development when measured capacity utilization, scheduling, or inventory constraints offer a material opportunity that outweighs implementation effort. Neither architecture is inherently superior for every molecule or program.
Record the evidence supporting the choice, the main uncertainty, and the conditions that would change the decision. Higher annual demand, a more expensive resin, altered feed concentration, or a longer campaign may justify revisiting an earlier conclusion. Retain a consistent baseline so later comparisons remain meaningful.
Review candidate antibody affinity materials and discuss purification requirements through protein purification services. If evaluating a specialized multi-column configuration, explicitly confirm equipment and development scope rather than inferring that a general service description covers the architecture. Match the technical proposal to demonstrated capabilities and the evidence your process requires.
Compare operator workload as well as automated run duration. Setup verification, buffer connections, sampling, troubleshooting, and data review may change substantially even when the instrument performs switching automatically. Include training and maintenance requirements in the readiness assessment. A well-understood conventional method can be preferable during an early program, while a later increase in demand may justify more complex operation. Record that stage-dependent rationale rather than treating the initial architecture choice as an irreversible technical judgment.
Assess the right chromatography architecture
Share the baseline capacity, cycle schedule, feed supply, annual demand, and product-quality requirements to identify whether an architecture comparison is warranted.
References
- Schmitz F, Minceva M, Kampmann M. Comparison of batch and continuous multi-column capture of monoclonal antibodies with convective diffusive membrane adsorbers. Journal of Chromatography A. 2024;1732:465201. doi:10.1016/j.chroma.2024.465201
- Carta G, Jungbauer A. Protein Chromatography: Process Development and Scale-Up. 2nd edition. Wiley-VCH; 2020. doi:10.1002/9783527824045
- Silva TC, Isaksson M, Nilsson B, Eppink M, Ottens M. Optimization of multi‐column chromatography for capture and polishing at high protein load. Biotechnology Progress. 2025;41(5):e70047. doi:10.1002/btpr.70047
