Define harvest as a handover of quantity and condition

A harvest campaign does not begin at a pump. It begins with a defined transfer from cultivation: product location, liquid volume, cells and solids, analytical product content, temperature condition, release status, and earliest and latest processing times. These determine whether the next operation has sufficient capacity and receives suitable material in time. Equipment can provide the required litre throughput yet remain unsuitable if product loses an essential property during waiting.

The case below concerns a dissolved extracellular protein. Cells are first removed, followed by concentration and diafiltration, then one capture step. This is a deliberately bounded process segment, not a complete medicinal-product process or universal purification platform. Intracellular products, harvested cells or insoluble inclusion bodies would require different material assignments from the outset. Understanding product location and separation sequence is therefore a prerequisite.

The balance boundary extends from available culture volume to capture pool. Each transfer carries product mass and liquid volume separately. Rising concentration does not automatically mean product gain; increasing pool volume does not automatically mean product loss. Flushes, samples and unrecovered residual quantities need explicit destinations. Processing, waiting, preparation and handover are distinguished on the timeline.

A process-specific handover sheet should contain the measurements actually needed. A universal form containing many irrelevant tests does not improve sizing. What matters is which property determines the next separation step or allowable hold, and how reliably that property has been established in the transferred material.

One continuous synthetic dataset

The starting harvest is 2,000 L at 2.0 g/L of target product, or 4,000 g. Every number in this case is an assigned teaching assumption. Clarification recovers 96% of product mass. Its transferred pool remains at 2,000 L through an explicitly included volume make-up and rinse balance; this does not imply that every real clarification preserves volume. The case therefore contains 3,840 g, or 1.92 g/L, after clarification.

Ultrafiltration, UF, concentrates this pool to 200 L. Product sieving is small but deliberately non-zero at 0.001. Diafiltration then proceeds at a constant 200 L for five diavolumes. Applying the same product sieving to both stages is a model simplification. A real dependence on concentration or composition would require separate measurement.

After UF and diafiltration, DF, 3,812.0602 g remains in 200 L, corresponding to 19.0603 g/L. A hypothetical 95% capture recovery gives 3,621.4572 g. Overall recovery across the defined segment is consequently 90.5364%. It is neither 96% + 99% + 95% nor an arithmetic average of step yields. Each step refers to its own incoming mass.

The executable Python file, membrane alternatives, capture cycles and overall results describe the same case. A changed UF recovery can therefore be traced into the final chromatography cycle instead of comparing disconnected example numbers.

Understand clarification as both separation and scheduling

Clarification begins by identifying the product-bearing fraction. A dissolved protein should remain in clarified liquid, whereas a cell product requires a different assessment of the solids. A visually clear sample does not fully describe clarification performance. Residual particles, colloids, soluble impurities and concentration distribution may matter more to the next operation than appearance.

Clarification in the case takes four hours from availability of the complete harvest, implying an average liquid throughput of 500 L/h. This is a teaching schedule assumption, not the capacity of a named centrifuge or depth filter. During processing, some material remains upstream while some has passed the separation. Early clarified fractions may therefore wait longer for common downstream processing than the last fraction.

A continuous liquid stream and a subsequently pooled batch have different age distributions. An initial conservative assessment can consider the full elapsed time since harvest availability. More detailed design models the collection and mixing history. An average hold must not conceal a potentially important older fraction. Its relevance depends on product stability and the defined quality question.

Assumed clarification loss is 160 g. A real study would determine how much resides in solids, residual liquid, adsorption or samples. A single 96% factor is transparent in a teaching case but does not diagnose the loss. Improvement requires locating it and supporting that allocation with an appropriate balance or recovery investigation.

Assess holding as a documented material history

Allowable hold time is not a universal property of proteins. It depends on product form, matrix, concentration, temperature history, contact material and the properties actually required. A literature example may explain a mechanism but cannot authorise storage of a different harvest. This also applies to the common assumption that a chilled sample remains stable indefinitely.

The planning exercise imposes a purely hypothetical limit of 30 hours between harvest availability and completion of capture. It is a constraint of the example, explicitly not an empirical stability value. A schedule exceeding it is rejected in this case. A schedule meeting it is feasible only with respect to that assigned condition; it does not demonstrate actual product stability.

A development study exposes samples to relevant time and condition histories and compares them using appropriate assays. A single final-storage result does not necessarily reveal when or why a change occurred. Subsequent separability may itself be a relevant response: unchanged total titre does not exclude changed aggregation or binding behaviour.

Documentation should identify start and end events precisely. Does the hold begin at the end of cultivation, at clarification completion, or after reaching a specified temperature? Does it end when loading starts or after the last fraction is processed? This definition may affect the result more than a small throughput adjustment. Timestamps and material identity are consequently part of the technical evidence.

Calculate the complete sequence for two alternatives

In the shared sequence, clarification starts at hour 0 and ends at hour 4. Transfer to UF/DF takes another 0.5 h. A 12 m² membrane requires 5.4012 h of filtration under the assigned flux model. Preparation adds 0.75 h, flushing 0.5 h and recovery 0.25 h. These activities are deliberately treated as sequential. The UF/DF segment therefore takes 6.9012 h and finishes at overall hour 11.4012.

A 20 L capture column processes the resulting pool in six cycles. Including one hour of preparation and every recurring auxiliary phase, capture takes 13.7 h. The campaign consequently ends at hour 25.1012, which is 4.8988 h before the hypothetical 30-hour limit. This is calculated distance from an assigned constraint, not a statistically established safety buffer.

A 10 L column needs twelve cycles. Capture takes 26.4 h and the full campaign ends at hour 37.8012. Although the smaller column can technically receive each individual portion, that combination fails the assigned time requirement. Demonstrating enough resin capacity through arbitrarily many cycles is therefore not complete sizing.

This sequence is deliberately a serial baseline. A real facility might prepare equipment in parallel or transfer partial pools earlier. Such improvements are valid if material flow, equipment state, personnel and holding conditions permit them. Overlapping bars on a diagram do not save time when the tasks require the same exclusive resource. The detailed cycle calculation and membrane sizing explain the assumptions.

Serial campaign with 12 m² UF/DF area
ActivityStart (h)Finish (h)
Clarification04
Transfer44.5
Complete UF/DF occupancy4.511.4012
Capture with 20 L resin11.401225.1012
Alternative: capture with 10 L resin11.401237.8012

Track buffers, pool volumes and auxiliary streams

Five diavolumes at a constant 200 L require 1,000 L of fresh exchange buffer in the idealised DF step. Previous concentration removes 1,800 L, and DF produces another 1,000 L of permeate. Total permeate handling is therefore 2,800 L. This is neither initial harvest volume nor final product-pool volume. Waste collection or weighing arrangements must accommodate the actual flow profile.

The capture case assigns 17 column volumes of buffers and process solutions per cycle. With 20 L of resin and six cycles, this totals 2,040 L. Elution uses three column volumes per cycle, giving 360 L of combined eluate. With 10 L of resin and twelve cycles, both totals are identical under the assigned proportional assumptions. The smaller column therefore does not save total recurring buffer in this case, although its individual container loads are smaller.

Preparation flushes, system dead volume and discarded transition fractions are not automatically fully represented. A real consumption plan includes them explicitly. Direct buffer provision, on-site dilution and repeated preparation in separate vessels change readiness time and equipment demand. Summing a recipe does not establish timely availability or appropriate quality at the use point.

Capture pool volume can exceed the concentrated UF/DF volume. Here that follows from elution volumes and is not an accounting error. A subsequent concentration would be another operation with its own recovery, time and membrane area. It is not silently included at no cost in this bounded campaign.

Locate losses rather than comparing percentages alone

The initial 4,000 g is distributed into 160 g clarification loss, 8.8318 g UF product passage, 19.1080 g DF product passage, 190.6030 g capture loss and 3,621.4572 g recovered product. These quantities close apart from rounding. Capture loss is an assigned aggregate factor, whereas membrane losses follow an explicitly derived passage balance. A real analysis should preserve this distinction in evidence quality.

Improving capture recovery from 95% to 97% would apply to its 3,812.0602 g inlet and recover another 76.2412 g. It would not simply equal 2% of the original 4,000 g. The additional amount must then be weighed against extra time, buffer demand or changed purity. Improving an isolated number can worsen the complete process.

A larger rinse may recover residual product while increasing the volume requiring treatment. A narrower pool cut may improve purity while reducing recovery. Faster filtration may save time but alter product passage. Every proposed change should therefore be evaluated within the same continuous mass and time calculation.

If loss has not been located, record it as an unexplained balance residual. That is more useful than inventing an explanation. Targeted samples from retentate, permeate, wash fractions and residual inventories can make the next experiment identify the missing material route.

Combine portions through their masses and states

When several harvest fractions are combined, pool concentration follows from total product mass and total volume. A separate example combines 70 L at 3 g/L with 30 L at 1 g/L. Product masses are 210 and 30 g. Assuming additive volumes, the 100 L pool contains 240 g, or 2.4 g/L. The unweighted concentration mean of 2 g/L would misdescribe the handover.

A product characteristic may need another weighting. If the two fractions contain 90 and 70 per cent of a species defined on a product-mass basis, their species quantities are 189 and 21 g. The combined fraction is 210/240 = 87.5 per cent. Volume weighting, 0.7 × 90 + 0.3 × 70 = 84 per cent, would be wrong because the fraction refers to product mass rather than liquid volume. This calculation is appropriate only for a genuinely additive species balance.

Not every quality measure can be averaged this way. Measured activity may depend on interactions or conditions; an individual requirement violation is not erased by a favourable pooled average. The numerical example therefore does not authorise arbitrary pooling. It identifies the reference definition needed before separate fractions can support a common quantity or property statement.

Check overlapping harvest through intermediate inventory

Under suitable conditions, downstream processing can start before harvest collection is complete. Calculate intermediate inventory over time. In a separate model, harvest supplies 500 L/h for four hours. A following step processes 600 L/h once sufficient feed is available. Starting immediately with an empty intermediate vessel cannot sustain those constant rates: outflow exceeds inflow and would require negative inventory.

Waiting one hour creates an initial 500 L buffer. Over the next three hours it decreases by 100 L/h to 200 L. After harvest finishes, the remaining 200 L takes one third of an hour to process. Completion is therefore 4.333 h after harvest begins. A fully serial arrangement takes four hours of collection plus 2,000/600 = 3.333 h of processing, so the idealised overlap substantially shortens the endpoint.

The calculation assumes a suitable intermediate vessel, unchanged processing capability under continuous feed and an acceptable material history. Maximum intermediate inventory is 500 L here, not the entire 2,000 L harvest. Individual portions nevertheless have different residence times. Actual implementation also needs interruptions, transfer inventory and end conditions examined. The inventory calculation tests whether a drawn overlap is physically compatible with material flow.

Prioritise losses by the largest avoidable contribution

Percentage losses in different steps usually use different incoming quantities. In the shared case, clarification loses four per cent of 4,000 g, or 160 g. UF/DF then reduces the remaining 3,840 g to 3,812.0602 g, losing 27.9398 g. Capture subsequently loses five per cent of its own input, or 190.6030 g. Simply adding four and five per cent therefore does not describe total process loss.

The final pool contains 3,621.4572 g. Relative to 4,000 g initially, total loss is 378.5428 g. The three absolute losses sum to the same value. This identifies where further investigation might affect a larger quantity, but does not establish which loss is technically avoidable. A small residual loss may arise from an essential separation, while a larger one may respond to improved recovery.

Prioritisation combines lost mass, plausible cause, existing evidence and the effort of a discriminating measurement. Re-evaluate the whole sequence after a change: more product after clarification may require an additional capture load. Higher recovery may also carry more impurities into a later step. The objective is usable final product meeting its requirements, not the highest isolated recovery from every individual unit.

Challenge the design with counterquestions

Question 1: A larger membrane completes UF/DF one hour earlier. Does the campaign automatically finish one hour earlier? Only if that operation lies on the limiting sequence and the next stage is available. If the pool must wait for an occupied capture skid, the gain may merely increase waiting. The membrane may still provide another benefit, such as reducing maximum time before UF/DF completion, but that benefit must be named.

Question 2: Culture titre increases from 2.0 to 2.5 g/L at unchanged volume. What must be recalculated? Starting mass, step losses, membrane product concentration, possible flux changes, capture load, integer cycles and pool quantities. Unchanged litre throughput does not establish unchanged downstream capacity. Keeping synthetic flux unchanged is only possible as an additional assumption requiring verification.

Question 3: An alternative meets the 30-hour limit exactly. Is it robust? The deterministic calculation then provides no distance from the assigned constraint. Sensitivity to duration, quantity and observed variability must be investigated. An arbitrary percentage allowance does not replace that work. A justified reserve identifies the uncertainty driver and its actual effect.

Question 4: Real stability data are missing. May the schedule be declared suitable for the product? No. It establishes only a possible mass and timing sequence under explicit assumptions. The next development step is a hold study representative of the material and temperature history, together with verification of assigned separation performance.

Sources and further design work

MIT's biochemical processing material explains links between product objective, purity and recovery. Its process simulation and design lecture adds continuous quantity and capacity assessment. Historical prices and market figures are not used as current data here.

Continue the calculations with membrane area sizing and chromatography cycle planning. Shared equipment and multiple batches are considered in scheduling and bottleneck analysis. The data here are deliberately synthetic; unknown real stability and unmeasured separation performance are not replaced by borrowed literature numbers.

Terms in this chapter

30 terms explained in the glossary