Pilot Scale Monoclonal Antibody Production Biology Essay


This paper discusses the use of computer simulation tools for modelling a pilot scale monoclonal antibody production in a contract manufacturing organization (CMO). A batch process simulation software named SuperPro Designer v6.0 (Intelligen, 2005) is employed to model and perform the scheduling task for a 1000 L monoclonal antibody production as a base case model. In order to increase the plant throughput, optimization strategies are proposed with alternative process setups that utilize idle equipment in the plant. As such, the CMO is able to produce six more batches per year and hence reduce its payback time to approximately two years. This optimized strategy is next used as extended base case where uncertainty analysis is carried out with Monte Carlo simulation. From the analysis, it is determined that cell culture time in the bioreactors have the greatest impact in achieving the targeted plant throughput and profitability.

To enhance manufacturing efficiency and hence business competitiveness in the pharmaceutical industry, various design and optimization techniques have been developed in the past decades. Computer-aided process simulation (CAPS) is one of such tools that has gained good attention in recent years in improving manufacturing efficiency. It involves the use of computers to perform steady-state heat and mass balancing, as well as sizing and costing calculations for a process (Westerberg et al., 1979). Most often, it enables the identification of missing parameters and predicts the behavior of an integrated process under varying operating conditions. CAPS has been commonly used in the bulk and petrochemical industries since the late 1960's. However this tool is relatively new to other manufacturing industries. For instance, in biochemical production, the use of CAPS has only been reported since middle 1990s (Petrides, 1994; Petrides et al., 1995). More recently, CAPS were also being used in pharmaceutical production (Petrides et al., 2002a, Tan et al., 2006), specialty chemical manufacturing (Athimulam et al., 2006); food and beverage processing (Bon et al., 2010; Alshekhli et al., 2011), as well as water treatment processes (Petrides et al., 2001).

Lady using a tablet
Lady using a tablet


Essay Writers

Lady Using Tablet

Get your grade
or your money back

using our Essay Writing Service!

Essay Writing Service

More recently, the incorporation of economic analysis and debottlenecking functions into CAPS tools enable the process designers to identify economic "hot-spot" of a process at the early conceptual design stage. Various options can then be incorporated and evaluated with CAPS tools to reduce capital and/or operating costs in order to increase production throughput (Koulouris et al., 2000; Petrides et al., 2002b).

In this work, a pilot scale monoclonal antibody manufacturing process in a contract manufacturing organization (CMO) is modeled using SuperPro Designer v6.0 (Intelligen, 2005). To increase production throughput, different optimization strategies with alternative process setups are proposed and evaluated using SuperPro Designer. This optimized strategy is next used as an extended base case where uncertainty analysis is carried out with Monte Carlo simulation to quantify risks in meeting targeted plant throughput and profitability.


The production of monoclonal antibody (MAb) from mammalian cell culture is simulated as a base case model in SuperPro Designer, with the simulation flowsheet shown in Figure 1. The MAb production consists of upstream (which consists of Inoculum Preparation and Cell Culture sections) and downstream processing (consists of Recovery and Purification sections).

Supporting operations such as media preparation, pre-operation and post-operation steps, i. e. cleaning in place (CIP) and sterilization in place (SIP) are also considered in the model. Buffer preparation, however, is not modeled for simplicity. It is assumed that all buffers, cleaning and storage solutions (apart from that of the automated CIP cycles) are prepared in advance prior to the operation and stored in disposable bags.

The MAb production starts with inoculum preparation in two spinner flasks (in the Inoculum Preparation section), each of 1 L working volume. Once the desired cell density is achieved, the inoculum is transferred from the spinner flasks to the 5 L bioreactor (00.05.D001). The media solution for spinner flasks is transferred directly from the manufacturers' packaging, while the media solution for 5 L bioreactor is prepared in a stainless steel preparation tank (01.03.T001), together with the media for the subsequent bioreactors (in the Cell Culture section). The media solution is sterilized by passing through a 0.2 µm sterile filtration unit (01.15.F002/3). Media solution for 5 L bioreactor (00.05.D001) and 30 L bioreactor (02.05.D001) is collected in disposable bags and transported into the designated production area, while media for 200 L bioreactor (02.06.D001) and 1000 L bioreactor (02.07.D001) is transferred directly into the recipient bioreactor via the piping panel.

Lady using a tablet
Lady using a tablet


Writing Services

Lady Using Tablet

Always on Time

Marked to Standard

Order Now

Upon the completion of the transfer, the media solution is stored in the bioreactors at 4 °C until the start of cell culture. Storage at low temperature is made possible by utilizing glycol as the cooling agent. As the 1000 L production bioreactor is operated in fed-batch mode, only 500 L of media solution is transferred at the start of cell culture. Another 300 L of media solution is prepared separately in the preparation tank (01.03.T001) and fed into the bioreactor during the cell culture. The cell culture in the 5 L, 30 L and 200 L bioreactors takes 3 days each, whilst the 1000 L bioreactor takes 15 days. The volume of final cell culture broth is approximately 1000 L, containing about 1 kg of MAb.

Upon the completion of the cell culturing process, the content of the bioreactors are tranferred to the Recovery section. Biomass and other suspended compounds (denoted as impurities) in the culture broth are removed using a disposable depth filter system (POD-1). The depth filter area is estimated as 15.4 m2 and the filtration rate is set to 38.5 L/min. It is estimated that 5% of the MAb is lost into the solid waste stream during the filtration step. The clarified solution is directed to a stainless steel mobile process vessel (02.23.T001). However, due to its limited capacity, the vessel can only contain 200 L of the clarified solution, while the remaining 800 L is collected in two units of 500 L disposable bag. This is followed by 10-fold concentration and 5 times diafiltration of the clarified solution using a 4.5 m2 filtration system (02.16.D001), with filtrate flux set to 140 L/m2h. Permeate of the diafilter is pooled in the mobile process vessel, which is then transferred into disposable bags to be sent to the Purification section.

In the Purification section, Protein A chromatography is carried out in two cycles using a chromatography system of 45 cm-diameter column (02.09.D001). The operating parameters are set as follow:

Bed height: 20 cm (i.e. a bed volume (BV) of 31.81 L)

Resin binding capacity: 25 g of product/L of resin

Product is recovered in 5 BV's of low pH Buffer B

Product recovery yield: 90%

Linear velocity: 200 cm/h for all operations

Total buffer requirements:

Buffer A: 13 BV's (for equilibration and wash out unbound)

Buffer B: 10 BV's (for product elution and column regeneration)

Wash Buffer: 3 BV's (for intermediate column wash)

The eluant from Protein A chromatography is collected in a 100 L disposable bag and incubated for 45 minutes for virus inactivation. Subsequently, the protein solution is transferred into a 150 L stainless steel process tank (03.23.T001). Tris solution is added into the tank to adjust the pH value of the protein solution for neutralisation. The solution then undergone 5-fold concentration followed by 5 times diafiltration (03.16.D001; with 1.5 m2 filter area and filtrate flux of 140 L/m2.h). The concentrated protein solution from each cycle of Protein A chromatography is pooled in a temperature-controlled disposable bag system (ALG-1).

The next product purification step is flow-through mode anion exchange chromatography (IEX) using a chromatography system with 25 cm-column (03.09.D001). In flow-through mode, it is the impurities that bound to the resin instead of the product. Thus, the product is captured during loading operation and not during elution. The IEX operating parameters are set as follows:

Bed height: 10 cm (i.e. BV of 3.14 L)

Linear velocity: 200 cm/h for all operations

Total buffer requirements:

Buffer A: 13 BV's (for equilibration and wash out unbound)

Buffer B: 10 BV's (for elution and column regeneration)

The protein solution that flowed through the IEX chromatography column is then filtered using a nanofilter for removal of virus residues. This is followed by 5-fold concentration and 5 times diafiltration of the protein solution using a filtre (04.16.D001; with 0.42 m2 filter area and filtrate flux of 140 L/m2.h). Permeate of ultrafiltration is collected in a 150 L stainless steel process tank (04.23.T001) and transferred into a 20 L disposable bags for polishing step in cation exchange chromatography (CEX) .

Polishing step in the CEX is executed in four cycles using a chromatography system with 20 cm-diameter column (04.09.D001). The operating parameters are set as follow:

Bed height is 20 cm. Thus, the BV is 6.28 L

Resin binding capacity is 40 g of product/L of resin

Lady using a tablet
Lady using a tablet

This Essay is

a Student's Work

Lady Using Tablet

This essay has been submitted by a student. This is not an example of the work written by our professional essay writers.

Examples of our work

Product is recovered in 5 BV's of Buffer B

Product recovery yield is 90%

Linear velocity is 200 cm/h for all operations

Total buffer requirements:

Buffer A: 13 BV's (for equilibration and wash out unbound)

Buffer B: 10 BV's (for product elution and column regeneration)

The eluted solution is collected in the same process tank (04.23.T001) used in the preceding procedure. The solution is then concentrated 2-fold and diafiltered 5 times with product formulation buffer (i.e. PBS), also using the same filtration system (04.16.D001). However, different set of membrane cassettes are used for this procedure. The concentrated protein solution finally passed through a 0.2 µm sterile filter for bioburden reduction prior to bulk filling into an appropriate product container.

In this base case model, different approaches are used to represent the three major chromatography steps; pre-operation, operation and post-operation. This is particularly crucial for chromatography procedures with multiple cycles. Pre-operation and post-operation activities are done only once per batch and not repeated in each chromatography cycle. Thus, the only way to represent these activities is by using a separate unit procedure that utilizes the same equipment. For instance, procedure P-41 is dedicated for CEX operation that consist of four cycles per batch, while P-41a and P-41b each represents pre-operation and post-operation activities for CEX that are only performed once per batch. Another approach is to use multiple unit procedures to represent the multiple cycles and to include the pre-operation activities in the first cycle and the post-operation activities in the last cycle. This is considerably easy for small number of cycles like in this case; Protein A chromatography that has only two cycles (P-26 and P-31). Otherwise, this approach is rather complicated and may result in a complex flowsheet.

We next discussion the scheduling aspect of the batch process modeling. In the modeling framework of SuperPro Designer, scheduling of all operations in each unit procedure is set up based on the setup time, process time, and start time (Intelligen, 2005). Setup time is the preparation time required before an operation takes place. This involves equipment preparation or operation setup such as connection of spool pieces and hoses, as well as the transfer of material from one processing area to another. On the other hand, process time represents the actual processing duration of the operation, whereas start time denotes the beginning of the operation.

Note that process time for certain operations is dependant upon other operations of the same or different procedure. This inter-dependency is represented using the Master-Slave Relationship feature in SuperPro Designer (Intelligen, 2005). For example, the transfer in inoculum operation in procedure P-13 is set to follow the duration of transfer out operation in Procedure P-12. In this case, the transfer out operation in Procedure P-12 behaves as the Master operation, whilst transfer in inoculum operation in procedure P-13 behaves as the slave operation.

Note that in order to accurately represent the process schedule, the start time of certain operations are given a time shift, mainly to avoid overlapping of operations when utilizing the same equipment. For instance, CIP operation in procedure P-24 has a start time shift of 2.50 hours after the previous operation (Flushing) ends. This is to allow the completion of CIP 1000L operation in procedure P-13 that uses the same CIP Unit. Similar time shift approach is also applied for CIP operations in procedures P-33 and P-38. On the other hand, multiple use of process tank 04.23.T001 for procedures P-38 and P-42, as well as filtration system 04.16.D001 for procedures P-39 and P-43, result in a time shift of 4.5 hours for transfer out operation in procedure P-40. This time shift ensures that sufficient time is allocated to completely clean these two equipment prior to the start of the later procedures. Scheduling of the base case model is represented in the Gantt chart as shown in Figure 2.

From the base case simulation model, the plant batch time is determined as 679.72 hours. This is the time required from the preparation of inoculum in the spinner flasks to the final filtration of product in a single batch. A new batch is initiated every 634.67 hours, which corresponds to the minimum cycle time calculated by SuperPro Designer. The 1000 L bioreactor (02.07.D001) of procedure P-13 is identified as the scheduling bottleneck. On an annual basis, the plant processes a total of 12 batches, or a total of 7.62 kg of MAb. For every batch, the MAb recovery yield is determined as 64%. In other words, 640 g of MAb is recovered from the 1000 L bioreactor's culture broth.

Table 1 shows the amount of raw materials required for the MAb production. As shown. large amount of RO Water is being utilized for each batch, primarily for CIP operations. A total of 5,650 kg WFI is also used for CIP operation (i.e. as final rinse). Apart from the automated CIP operations, all buffers, cleaning and storage solutions are also prepared from WFI. However, the amount of WFI used for the make-up of these buffers and solutions are not shown in the table, but can be calculated from its mass composition, as defined in the stock mixtures databank.

To perform economic analysis of the MAb production, indicative values from the CMO are used. These include the cost of raw materials, consumables, utilities, waste disposals and labour, as well as other economic evaluation parameters such as depreciation period, income tax rate, etc. The total capital cost investment estimated by SuperPro Designer using its built-in economic evaluation model (Intelligen, 2005) is estimated as $36 million. This estimation is inline with the CMO's actual capital investment. The plant revenue is calculated based on the production rate of MAb in stream S-208 at a selling price of $1,500/gram. With a unit production cost of approximately $1,000/gram of the purified MAb, the project yields a payback time of 5.18 years, with a gross margin of 32.45 % and return on investment (ROI) of 19.32 %. The after-tax internal rate of return (IRR) is 12.27 % and the net present value (NPV) is $13 million (based on discount interest of 7%). As an incentive to promote the growth of biotechnology, the CMO is exempted from income tax for the first ten consecutive years the company derived statutory income from its business. Thus, the income tax rate of the model is set to 0 %.

Figure 3 shows the break down of the plant annual operating cost. The facility-dependent cost is the main contributor, accounting for 74 % of the operating cost. This is common for high value-added products that are manufactured in small quantities. Consumables are in the second position with 13 % of the total operating cost, which include disposable bags for storage of media and buffers, as well as chromatography resins and membrane filters that need to be replaced on a regular basis. Raw materials and laboratory/QA/QC components each contributes 5 % of the total cost, followed by labour cost at 3 %. In this case, the labour cost is set to $5.75 per hour for equipment operation, incorporating factors for fringe benefits, operating supplies cost, supervision cost and administration cost based on the built-in model in SuperPro Designer (Intelligen, 2005). On the contrary, the cost of labour for laboratory/QA/QC works is defined as 15 % of the total labour cost and is inclusive of the laboratory/QA/QC cost. The laboratory/QA/QC cost also covers detailed costing of all tests based on the defined frequency and cost per test, either on an annual or per batch basis. Tables 2 and 3 provide the sectional lists of laboratory/QA/QC tests considered in the base case model. A section named Facility is created in the simulation model to include the annual operating cost in maintaining the facility and utility systems.


In order to increase the plant throughput and profitability, two optimization strategies are proposed. These optimization strategies, however, are only limited to the existing plant and the equipment housed within it. This is because the CMO has no intention to purchase new equipment due to space constraint in the plant. Apart from the equipment used in the base case model, the following idle equipment, which is designed for the processing of smaller batch size (200 L), may be utilized in the optimized model.

Stainless steel preparation tank (01.04.T001) of 300 L working volume

Chromatography column (diameter = 10 cm, height = 50 cm)

Chromatography column (diameter = 30 cm, height = 50 cm)

Optimization Strategy 1 suggests an alternative setup for media preparation where the idle preparation tank 01.04.T001 is used in addition to preparation tank 01.03.T001. Instead of preparing the media for 5 L, 30 L, 200 L and 1000 L bioreactors in a single tank (01.03.T001), the media is prepared in both 01.03.T001 and 01.04.T001. This setup is illustrated in Figure 4. Preparation tank 01.04.T001 is assigned to prepare the media for 5 L, 30 L and 200 L bioreactors in procedure P-45, while preparation tank 01.03.T001 is dedicated to prepare media for 1000 L bioreactor in procedure P-4. This strategy allows the start of a new batch every 491.60 hours (approximately 6 days earlier than the base case), and subsequently increases the plant throughput to 18 batches per year. The simulation result indicates that 15.10 % increase in operating cost is observed, mainly due to more cleaning operations and the consumption of the 0.2 µm filters in the sterilization of media. Nevertheless, this increase is easily compensated by the significant increase in the number of batches processed per year. Moreover, this alternative setup could reduce the risk of contamination of media in the 1000 L bioreactor, since the media preparation is scheduled just before the transfer of inoculum. This is in contrast with the base case model where the media is prepared in advance and stored at 4 °C in the bioreactor for about 10 days before the transfer of inoculum. Figures 5 and 6 show the difference in the equipment occupancy for both models and the earlier start of a new batch.

Apart from reducing the minimum cycle time, the process can be further optimized using Strategy 2. In this strategy, the batch time is reduced by minimizing the number of cycles for cation exchange chromatography (CEX) in procedure P-41. This is made possible by using a larger column so that more products can be loaded in every cycle. Instead of four cycles in Optimization Strategy 1, procedure P-41 is carried out in only two cycles using a column of 10 cm larger in diameter. The comparison between the procedure setup for Optimization Strategies 1 and 2 is given in Table 4. When using a larger column, more chromatography resin is required to pack the column. Nonetheless, the resin is replaced less frequently as a consequence to the less number of cycles per batch. Furthermore, fewer buffers will be consumed for the procedure. These savings are reflected by a 1.04 % reduction in the operating cost as compared to Optimization Strategy 1 (see Table 5).

Table 5 shows the results of throughput analysis and economic evaluation for the simulation models of the base case, Optimization Strategies 1 and 2. By comparing the economic results, it is apparent that Optimized Strategy 2 with higher ROI and lower payback time would be the best solution to increase the CMO's plant throughput and profitability. This model, henceforth denoted as the extended base case, which are used as the basis for uncertainty analysis, as described in the following section.


The simulation model constructed using SuperPro Designer is of deterministic nature. This means that the model will provide reproducible outputs and does not consider the random variation of the inputs. For the simulation models discussed in earlier sections, average value has been used for the varying process input (e.g. cell culture time). In this section, uncertainty analysis is performed on the extended base case model using Monte Carlo simulation. The simulation quantifies the risks in meeting targeted plant throughput and profitability due to uncertainties in operational parameters and variability in the cost of raw materials and consumables.

Monte Carlo simulation is carried out by integrating SuperPro Designer with an Excel add-in application called Crystal Ball (Intelligen, 2005). The framework for integrating the two tools is shown in Figure 7. The probability distributions of the uncertain input parameters for Monte Carlo simulation are defined in Crystal Ball. When running the Crystal Ball application, random values for these parameters are generated using the Monte Carlo method according to their assigned distribution. For each simulation trial, the random values of the uncertain input parameters are sent to SuperPro Designer, which will then perform various calculations for the process flowsheet. This includes material and energy balances, scheduling and capacity utilization calculations, cost estimation and economic evaluation. The values of the output variables from SuperPro Designer are then sent back to Crystal Ball as forecasts. The input and output variables are linked to each other in an Excel spreadsheet using VBA scripts written in C language (Intelligen, 2005).

Figure 8 shows the flowsheet of the extended base case model, corresponds to the results of Optimization Strategy 2. As mentioned in earlier section, the plant processes a total of 18 batches of MAb per year. The simulation results also determine that the unit production cost is estimated as $770/kg MAb. Assuming the company has a production target of 14 successful batches per year, and an upper limit of $900/kg of MAb for the unit production cost, the above-mentioned analyses show that the company is able to meet its production and unit cost targets. However, it is beneficial to perform uncertainty analysis to assist the CMO in quantifying the risks of not meeting these targets.

The uncertainty analysis is focused on parameters that may have direct impact on plant throughput (i.e. number of batches per year) and profitability (i.e. unit production cost). The major contributors to the unit production cost are the costs of facility-dependent, consumable, raw material and laboratory/QA/QC. Since production is carried out in an existing plant, variation in facility-dependent cost is negligible. The same assumption goes to laboratory/QA/QC costs, which do not vary much when there are no major changes in the process.

Significant uncertainty, however, is expected on the cost of raw materials and consumables as they are highly associated with the supply and demand situation, as well as the world economics. When demand is more than supply, the material price will increase naturally, and vice versa. Furthermore, most raw materials and consumables used in the production at the CMO's site are imported. Fluctuation in world economics will therefore have significant impact on the cost of raw materials and consumables. Table 6 shows that 1X Media Solution is the most expensive raw material, which contributes 79.5 % of the raw materials cost. On the other hand, Virus Filter is the most expensive consumable, representing about 21.7 % of the consumables cost, followed by Protein A at 17.6 %, as illustrated in Table 7.

The plant annual throughput is determined by the minimum cycle time of the scheduling bottleneck (i.e. P-13). Hence, any process change that increase the cycle time of P-13 will result in lower level of annual batch production. In other words, any variability in the completion of P-13 will lead to uncertainty in the plant throughput. This variability is not limited to operations within P-13 alone, but also the variability in the various procedures upstream of P-13. Procedures P-3 and P-10 for example, poses uncertainty in their operations due to the variability in the skills of operators during manual equipment setup (5 L bioreactor and 30 L bioreactor respectively). The worst case would be cell culture contamination in these bioreactors, which will require a complete restart of the entire procedure.

Table 8 summarizes the input parameters chosen for the uncertainty analysis and their assumed probability distributions. Two forecast variables are considered in the simulation, i.e. the number of batches per year and the unit production cost of the MAb. These variables are chosen based on their significance in production planning and process economics. The output variables of the Monte Carlo simulation are quantified by mean, median, mode, variance, standard deviation and frequency distribution.

The simulation results (i.e. annual batch production and unit production cost) are presented in frequency distribution curves for the forecast variables. After 1,000 simulation trials, the distribution curves for the forecast variables are normally distributed. The Frequency Chart in Figure 9 reveals that the company is able to achieve its production target of minimum 14 successful batches per year with a certainty of 87.20 % (represented in blue bars of the curve). On the other hand, the Frequency Chart in Figure 10 shows that the certainty of meeting the company's unit production cost target of $ 900,000 is 87.10 %.

Apart from Frequency Charts, Sensitivity Charts provide overview on the variation of the forecast variables with respect to the uncertain parameters. This is very useful as it allows the company to identify which input parameter have the greatest impact on the plant throughput and profitability, and thus focus the effort to improve this parameter. The Sensitivity Charts for the number of batches per year and the unit production cost are given in Figures 11 and 12 respectively. Cell culture time in P-10 (i.e. 30 L bioreactor) has the greatest impact on both the number of batches per year (accounting for 65.4% among the factors), as well as the unit production cost (contributes to 58.6% among the factors). This is followed by cell culture time in P-3 (i.e. 5 L bioreactor). Cell culture time in P-13 (1000 L bioreactor) and 1X Media Solution cost also contribute to the unit production cost, but with much smaller percentage (3.6 % and 2.9 % respectively). Therefore, the company should focus its improvement efforts on the operation of 5 L and 30 L bioreactor to have a better certainty of meeting its throughput and profitability targets. Better understanding of the process, good process handling, well-trained operators and implementation of advanced automation can help to reduce the variability in the operation of these bioreactors.


This paper demonstrates how CAPS tools are used in modeling and optimizing a pilot scale production of monoclonal antibody. In base case simulation, SuperPro Designer is used to simulate and schedule the production process. Optimization strategies are then proposed and evaluated using the software in order to increase the plant throughput and profitability. Optimization Strategy 2 proves that even without the purchase of new process equipment, the annual throughout of the plant can be increased significantly by reducing the minimum cycle time. This is made possible by using alternative equipment setups and utilizing idle equipment available in the plant.

An uncertainty analysis study is then carried out by integrating SuperPro Designer with a Monte Carlo simulation software known as Crystal Ball. This quantifies the risk in meeting the CMO's target in terms of throughput and profitability. Uncertainties in operating parameters and variability in the cost of raw materials and consumables are taken into account in the simulation. From the analysis, it is determined that cell culture time in the bioreactors have the greatest impact in achieving the targeted plant throughput and profitability.