Model supporting bioprocess monitoring using

Abstract

A method is provided for monitoring a cell culture in a culture medium in a container (3) by means of a computer (2) having software (5) wherein the container (3) is connected to at least one sensor (6) for measuring the cell culture (5), the software (5) uses the spectral data measured by the at least one sensor (6) to create and operate a data model (8) to calculate specific process parameters of the cell culture, in which first a synthetic cell-free culture medium sample is measured with the at least one sensor (6); comprising different ranges of culture medium compounds and / or metabolites and / or products and / or process parameter combinations, and then real-time monitoring of target cell culture parameters is performed using the data model (8) trained by the software (5).

Description

Technical Field

[0001] The invention described herein discloses a method for monitoring cell cultures in a container using Raman spectroscopy supported by a synthetic data model via a computer. This invention relates to the technical field of biopharmaceutical processes. Background Technology

[0002] In the chemical and pharmaceutical industries, quality management focuses on improving and increasing the productivity of biochemical compound manufacturing. This requires the use of complex bioprocesses with integrated real-time monitoring within the production line. Online analysis can automate processes, thereby optimizing them by significantly saving time and materials. In the classic approach of using Raman spectroscopy to monitor cell cultures, several batches of cell cultures are required. During each cell culture, Raman spectra are acquired and samples are collected for reference measurements. The reference measurements are then correlated with the Raman spectra of at least three or more batches to obtain reliable monitoring results for new batches. To enhance process automation, data models are now used to analyze and / or compute Raman spectra. Constructing such Raman spectroscopy data models applicable to the relevant bioprocesses requires vast amounts of data, and generating this data is time-consuming and labor-intensive. Such experiments are costly in terms of consumables, the time of bioprocess experts / engineers, and resources. For example, applying Raman spectroscopy in cell cultures requires several cell cultures with settings identical or similar to those intended for monitoring, each requiring weeks or even months of work. Because building the first model requires several cultures, the model building phase can take several months. This step is challenging for the industry because it requires a significant upfront investment of time and money, with no immediate returns. Furthermore, if users want to make adjustments to their process and need to change cell culture media and / or cell lines, the previously developed model may no longer be applicable, as typical data models used in Raman spectroscopy are only reliable for the same cell lines, the same cell culture media, and at similar bioreactor volume scales. Summary of the Invention

[0003] Therefore, the objective of this patent application is to find an improved method for monitoring cell cultures in containers / bioreactors that overcomes the known limitations of the prior art.

[0004] This task is addressed by a method for monitoring cell cultures in a culture medium within a container using computer software. The container is connected to at least one sensor that measures the cell cultures. The software uses spectral data measured by the at least one sensor to create and run a data model to calculate specific process parameters of the cell cultures. The data model is first constructed and trained by measuring a synthetic cell-free culture medium sample with the at least one sensor, encompassing a design space of different ranges of culture medium compounds and / or metabolites and / or products and / or process parameter combinations. The software-trained data model is then used for real-time monitoring of target cell culture parameters. The idea behind using a synthetic model is to explore the design space in the cell culture medium before cell culture begins. The principle is to synthetically reproduce the design space of the target parameters we want to monitor during cell culture in a cell-free environment, without requiring multiple cell culture runs to create the model.

[0005] The range of various components to be modeled in the cell culture medium intended for cell culture is defined. This also includes experimental design to map the relevant design space, thereby generating combinations of various components in the culture medium. Raman spectra are acquired for each condition, and a ready-to-use model can be built for the user, who can then run their first cell culture with a Raman probe that provides online and real-time quantitative data. Using synthetic samples significantly reduces the data acquisition time for obtaining the first model; in this case, data generation takes only a few days or a week. It also achieves cost reduction and portability to other cell culture media, cell lines, and potential volume scales.

[0006] Preferred uses of this process include, for example, but not limited to: • By using experimental design techniques, the relevant design space of various combinations of culture medium components within different ranges can be mapped.

[0007] • Use at least one Raman spectroscopy probe to provide quantitative real-time Raman data for each relevant culture medium condition.

[0008] • Use a bioreactor as a container, or a beaker as a container, or any vessel suitable for cell culture as a container. • Or use any liquid vessel in which cell culture medium can be measured as a container.

[0009] • Data modeling is performed by measuring the relevant design space in-line within a bioreactor or container, or by measuring the relevant design space at-line within a container using experimental design techniques, or a combination of both.

[0010] • For measurements in bioreactors, a culture medium containing no or only low concentrations of the component to be monitored is prepared, sterilized, and placed in a sterile bioreactor equipped with a Raman probe. The medium is kept under agitation, aeration, and temperature control using the same set points as typical cell culture bioprocesses, and the target metabolite can be added to create a range of concentrations.

[0011] • For measurements using experimental design techniques in containers, various combinations of culture medium compounds and / or metabolites and / or products and / or process parameters (such as glucose and lactate) to be monitored are created in the culture medium to map to combinations of concentrations encountered in actual cell culture processes. Culture medium samples are prepared by adding different concentrations of glucose and lactate solutions to each culture medium sample mixture. Products can be, for example, biological or cellular products such as proteins, antibodies, mRNA, etc.

[0012] • The model construction process was carried out under cell-free conditions.

[0013] Another solution to this task is an automated system for monitoring cell cultures in a culture medium, comprising a container equipped with at least one Raman sensor to measure cell cultures, a computer connected to the at least one sensor, and software running on the computer, the software having a data interface for managing the connection to the at least one sensor and providing a data model, the system being configured to perform the aforementioned method.

[0014] Preferred further improvements to the automation system include, for example, but not limited to: • The at least one sensor is a Raman spectroscopy probe.

[0015] • The container is a bioreactor or beaker, or any vessel suitable for cell culture, or any liquid vessel in which cell culture medium can be measured.

[0016] • Use offline desktop software for data acquisition and / or data modeling, such as SIMCA or Matlab.

[0017] Compared with known prior art, the method of the present invention has several advantages. These advantages include: • Faster model building for quick first robust monitoring and less time required to build the model.

[0018] • The data model is not dependent on a specific process, culture medium, scale, or cell line.

[0019] • The model exhibits better relevance and robustness for specific components.

[0020] • Reduce operation time, consumables, resources and equipment fixed costs. Detailed Implementation

[0021] The method according to the invention and the automated system 1 including software 5, and its functionally advantageous improvements, will now be described in more detail using at least one preferred exemplary embodiment and with reference to the accompanying drawings. In the drawings, corresponding elements are labeled with the same reference numerals.

[0022] The attached diagram shows: Figure 1 A schematic overview of the Raman spectroscopy system used. Figure 2 A general diagram illustrating the process of building and applying data models. Figure 3 : Measurement apparatus for range and DoE (Design of Experiments) Figure 4 This map shows the concentrations of glucose and lactate encountered during cell culture, the composition of the DoE sample, and the range of glucose and lactate levels. Figure 5 This image shows the concentrations of glucose and lactate encountered during cell culture, as well as the composition of the DoE sample. Figure 6 The results table for glucose and lactate shows the calibration set (synthetic data) on the left and the validation set (cell culture data) on the right. Figure 7 Comparison of glucose predictions from the constructed data model with measured reference values. Figure 8 Comparison of lactate predictions from the constructed data model with measured reference values. Figure 9 A comparison of the results from existing classical cell culture models and the synthetic model method of this invention, focusing on glucose parameters. Figure 10a A schematic comparison between model building methods and monitoring processes using existing technologies. Figure 10b A schematic comparison between the model construction method and the monitoring process using synthetic methods. Figure 1An example of a Raman spectroscopy system 1 used in this invention is shown. The system includes a bioreactor 3 itself (containing biomass containing cell cultures), its control unit 2, a Raman probe 6 connected to the bioreactor 3, and software 5 running by the control unit 2. This software calculates specific cell culture parameters using at least one specific data model 8 by analyzing real-time spectral data transmitted from the Raman probe 6 to the control unit 2. The control unit 2 is preferably a standard computer suitable for controlling the bioreactor 3. Another option is a microcontroller or processor integrated into an embedded device associated with the bioreactor 3. It can also be a standard or industrial personal computer or server or any other suitable device, especially where the local control unit 2 provides the data model 8 itself, as this requires higher processing power as typically provided by a microcontroller. In another preferred embodiment, the data model 8 is provided by a suitable standalone computer located at a remote location via a data network (using cloud-based services) or storage device, which has the same or different software.

[0023] Figure 2 A general schematic diagram illustrates the process for constructing a data model and applying it to its use in cell culture monitoring. Two types of measurement methods are preferably used here: 1. Perform range measurements online in a bioreactor or container.

[0024] 2. Use the DoE method to perform measurements in, for example, containers or sample tubes.

[0025] One or both of these methods can be used for model building. Raman spectra from cell-free culture medium formulations are acquired to build the data model. The data model described here is then applied to monitor cell culture.

[0026] Figure 3 The apparatus for range measurements and DoE measurements performed in a bioreactor or container 3 is shown. In the following sections, these two methods, along with the subsequent model-building steps, will be explained in more detail using corresponding preferred working embodiments.

[0027] 1. Range measurement in bioreactor 3: A cell culture medium free of the target molecule was prepared, sterilized, and placed in a sterile bioreactor vessel 3 equipped with a Raman probe 6. The medium was kept under agitation, aeration, and temperature control, using the same or similar set points as typical cell culture operations. A sterile bottle containing glucose or lactate was then inserted into the bioreactor and connected to a pump. The glucose or lactate solution was then slowly added to vessel 3 in a linear ramp manner, and Raman spectra were automatically acquired at frequent time intervals. Samples were collected during the experiment for reference measurements of glucose and lactate. The values ​​of glucose and lactate were then determined using, for example, linear regression, such that each Raman spectrum had a concentration matching the desired molecule (see also...). Figure 4 When the desired range for a specific metabolite is reached, the pump is stopped, and data is collected to build the model. The desired range is determined based on the process to be monitored in cell culture, with the goal of making the synthetic range larger than the process range: for example, if the glucose range in cell culture is intended to be 2.5–8 g / L, then the range for Model 8 could be 0–10 g / L. The pump rate and Raman acquisition frequency are selected to allow for the collection of a large number of data points. Here, more than 100 points are collected for each metabolite, running fully automated for one or several days, meaning that measurements do not require human intervention. A single run can be performed for each metabolite to create a complete dataset.

[0028] 2. Use DoE measurement in container 3: The DoE (Distribution of Evaluators) approach aims to formulate mixtures of different metabolites (such as glucose and lactate) in cell culture media to map to combinations of concentrations encountered in actual processes. Samples are prepared in cell culture media without the components to be monitored, and different concentrations of glucose and lactate solutions are added to each sample mixture. The DoE is constructed to select the composition of each mixture within a desired range. The desired range is determined based on the process to be monitored in cell culture, with the goal of making the synthetic sample mix larger than the process range. For example, Figure 5 The graphs illustrate the concentrations of glucose and lactate encountered during cell culture. The composition of the synthetic sample mixture was chosen to map... Figure 5 The concentration combinations shown are illustrated in the figure. The figure plots cell culture data points (ZN1-6), with a cross representing a sample from a cell culture run, showing the glucose and lactate concentrations. Dots represent the concentrations selected for the synthetic sample mixtures. In this embodiment, 13 sample mixtures were selected to map the design space. Each of the 13 sample mixtures was prepared in triplicate and Raman measured using Raman probe 6, yielding 13 × 3 Raman spectra. Reference values ​​were measured for each sample to construct the corresponding data model.

[0029] The data acquired using the Raman probe 6 was preprocessed and exported using preprocessing methods, for example, in the following order: 1) Perform SNV (normalization) calculations on the water region of the spectrum. 2) Savitzky-Golay first derivative, second-order smooth polynomial, and five-point window (corresponding to 15 cm) -1 ) 3) Select the Raman shift as follows: a. For glucose: 555-708; 855-1254; 1338-1380; 1596-1923; 3240-3393; 3432-3672 (cm) -1 ) b. For lactate: 423-456; 1008-1095; 1212-1401; 1518-1596; 2800-3000 (cm) -1 ) Then, the exported file (e.g., a .csv file) containing the preprocessed spectral data and reference values ​​is imported into a suitable analysis software module (Bio4C is used here). ® Use PAT Chemometric Expert to build at least one PLS data model. Once built, export the data model as a file and then upload it to another software module (Bio4C used here). ® PAT Raman Software 5) was used to predict cell culture batches (ZN2-10 batches).

[0030] The results regarding glucose are shown in Figure 6 In the table above, the left side represents the calibration set (synthetic data), and the right side represents the validation set (cell culture data). The abbreviations have the following meanings: LV: The number of latent variables in the PLS model.

[0031] N: Number of calibration samples.

[0032] Calibration range: The concentration range of the analyte.

[0033] RMSE: Root Mean Square Error of Calibration (C), Cross-Validation (CV), and Prediction (P) The results regarding lactate are presented in a similar manner. Figure 6 In the table below. Figure 7 and Figure 8 The data from the constructed data model 8 shows a comparison with measured reference values ​​for glucose (…). Figure 7 ) and lactate ( Figure 8The corresponding prediction results.

[0034] Figure 9 A comparison is shown between the results of constructing a synthetic model 8 from a known classical cell culture model 8 with the results of the present invention, for glucose parameters. For this method, a ZN-8 batch is used as an example. Figure 9 The results shown clearly demonstrate that the constructed synthetic data model 8 outperforms (by approximately 3 times) the real cell culture model used for monitoring glucose parameters. Figure 10a and 10b A comparative diagram of the two methods is also provided. Both diagrams illustrate the model construction methods and monitoring processes. Figure 10a The previous method required a process time of three to six months. The performance results demonstrated involved cell density and major metabolites, including glucose and lactate. The method of this invention, using synthetic data model 8, requires only one to two weeks of process time. Its superior performance has been confirmed for the glucose parameter. Data for the lactate parameter are also promising, but its superiority has not yet been numerically determined.

[0035] Three more preferred working embodiments will be disclosed next.

[0036] In the first embodiment (also known as "homemade model" 8), a specific Raman PAT platform user prepared some samples using the following methods: • Fresh culture medium • A series of components or products to be monitored (glucose, lactate, etc.) • As part of DoE, several components can be added to a sample at the same time to take into account the cross-interactions between components.

[0037] Raman spectra of each sample were acquired and preprocessed using different algorithms, such as normalization, Savitzky-Golay derivative, and spectral selection. A chemometric data model 8 was then constructed for each component to be monitored. The preprocessing used for model 8 was highly specific and required optimization for each component. The concentrations of these components were then measured in real-time using data model 8 in cell cultures using the same culture medium. Data model 8 can be used with different culture media and different cell lines, but its performance remains to be confirmed.

[0038] The second embodiment is an embedded model, which operates on the same principle as the self-made model 8 described above. The difference is that this data model 8 is built in a factory and embedded into the system to be delivered to the user. In this way, the user can directly use the data model 8 without having to build a model beforehand or collect data from it. In this case, the data model 8 can be built using the same culture medium used on the instrument, or it can utilize data collected from different Raman analyzers 1 to ensure sufficient robustness.

[0039] The third embodiment uses a large database from different Raman analyzers 1, whose software or database is embedded or cloud-based. In this embodiment, Raman spectra are acquired on different culture media on different Raman analyzers 1. All data are used to construct a universal model that can be used on any culture medium. Each synthetic formulation can be added to the database to make model 8 increasingly robust. The performance of the model must be tested to check whether the model is effective enough for the selected culture media. The large database can be installed at the plant used and updated during maintenance; it can also be shared via the cloud and updated in real time by other users.

[0040] Software 5 can automatically select data from a large database to create a corresponding data model 8 based on information about the process to be monitored (e.g., parameters such as the culture medium used and the type of process).

[0041] List of reference numerals 1. Raman spectroscopy system 2 Computers 3. Container (Bioreactor) 4. User Interface 5 software 6 Raman probe 7 users 8. Data Model.

Claims

1. A method for monitoring a cell culture in a culture medium in a vessel (3) by means of a computer (2) with software (5), wherein the vessel (3) is connected to at least one sensor (6) measuring the cell culture (5), the software (5) creates and runs a data model (8) using spectroscopic data measured by the at least one sensor (6) to calculate specific process parameters of the cell culture, wherein the data model (8) is first constructed and trained by measuring synthetic cell-free culture medium samples with the at least one sensor (6) including a design space of different ranges of combinations of medium compounds and / or metabolites and / or products and / or process parameters, and then the real-time monitoring of the target cell culture parameters is performed using the data model (8) trained by the software (5).

2. The method according to claim 1, wherein the relevant design space of different ranges of combinations of various medium compounds and / or metabolites and / or products and / or process parameters is mapped out by using experimental design means.

3. The method according to any of the preceding claims, wherein the at least one sensor (6) is a Raman probe and is used to provide real-time data for each relevant bioprocess condition.

4. The method according to any of the preceding claims, wherein a bioreactor (3) is used as vessel (3), or a beaker is used as vessel (3), or any suitable vessel for cell culture is used as vessel (3), or any liquid vessel in which cell culture fluid can be measured is used as vessel (3).

5. The method according to any of the preceding claims, wherein the data model (8) construction is performed by measuring the relevant design space online in a bioreactor or vessel (3) or offline in the vessel (3) using experimental design means or a combination of both.

6. The method according to claim 5, wherein for the measurement in a bioreactor (3), a medium is prepared which does not contain or contains only low concentrations of the components to be monitored.

7. The method according to claim 6, wherein the medium is sterilized and placed in a sterile bioreactor equipped with a Raman probe (6), wherein the medium is brought under agitation, aeration and temperature control using the same or similar set points as in a typical cell culture bioprocess.

8. The method according to claim 5, wherein for the measurement using experimental design means in the vessel (3), various combinations of medium compounds and / or metabolites and / or products and / or process parameters to be monitored in the culture medium, such as glucose and lactate, are created to map out the concentration combinations encountered in an actual cell culture bioprocess, wherein the culture samples are prepared by adding different concentrations of glucose and lactate solutions to each culture sample mixture.

9. The method according to any of the preceding claims, wherein the model construction process is performed under cell-free conditions.

10. An automated system for monitoring a cell culture in a culture medium, the system comprising a vessel equipped with at least one sensor (6) to measure a cell culture process parameter, a computer (2) connected to the at least one sensor (6) and a software (5) running on the computer (2), the software (5) having a data interface for managing the connection to the at least one sensor (6) and providing a data model (8), the system being configured to perform one of the preceding claims.

11. The automated system according to claim 10, wherein the at least one sensor (6) is a Raman spectroscopy probe.

12. The automated system according to claim 10 or claim 11, wherein the vessel (3) is a bioreactor (3) or a beaker, or any suitable vessel for cell culture, or any liquid vessel in which a cell culture liquid can be measured.

13. The automated system according to any one of claims 10 to 12, wherein an offline desktop software for data acquisition and / or data modeling is used, such as SIMCA, Matlab as software.

Images