Model supported bioprocess monitoring using raman spectroscopy

Abstract

A method to monitor cell cultures in a medium in a vessel (3) via a computer (2) with a software (5), where the vessel (3) is connected to at least one sensor (6) measuring the cell cultures (5), a data model (8) is created and ran by the software (5) using spectral data measured by the at least one sensor (6) to calculate specific process parameters of the cell cultures, wherein the data model is first built and trained by measuring synthetic, cell-free media samples with the at least one sensor (6) including a design space of different ranges of combined media compounds and / or metabolites and / or products and / or process parameters before using the trained data model (8) by the software (5) to perform real-time monitoring of the target cell culture parameters.

Description

[0001] Model supported Bioprocess Monitoring using Raman spectroscopy

[0002] The hereby described invention discloses a method to monitor cell cultures in a medium in a vessel via a computer using Raman spectroscopy supported with a synthetic data model.

[0003] Technical Field

[0004] The invention deals with the technological area of a biopharmaceutical process.

[0005] Background and description of the prior art

[0006] The chemical and pharmaceutical industry's quality approach is focused on improving and increasing productivity in the manufacture of biochemical compounds. This requires the use of complex bioprocesses with real-time monitoring integrated within the production line. In-line analysis can enable the process automation, thus optimizing it by a significant saving of time and materials. In the classic approach to monitor cell cultures using Raman spectrocospy several batches of cell cultures need to be ran. During each cell culture, Raman spectra are collected along with samples to make a reference measurement. The reference measurements are then associated to the Raman spectra for at least three or more batches to have reliable monitoring results on a new batch. To enhance the process automation, data models are nowadays applied to analyze and / or calculate the Raman spectra. To build such a data model for Raman spectroscopy which can be applied in respective bioprocesses, a lot of data are required, which takes long experiments to generate. Such experiments are expensive in terms of consumables, bioprocess expert(s) / engineer(s) time and resources. For example, to apply Raman spectroscopy in cell culture, several cell cultures with the same or similar setup as the one to be used in monitoring are required, and each cell culture is a work of several weeks or even months. As several cultures are required to build a first model, the phase of model building takes several months. This step is an issue for the industry because it requires a lot of time and money which needs to be beforehand without getting any advantages at this point. Also, if the user wants to make some adjustments to their process and needs to change the cell culture media and / or cell-line the previously developed models are most likely not applicable anymore, because the typical data model used in Raman spectroscopy is reliable only on the same cell-line, same cell culture media and in a similar bioreactor volume scale.

[0007] Summary of the invention

[0008] The task of this patent application is therefore to find an improved method to monitor cell cultures in vessels / bioreactors which can overcome the known limitations of the prior art.

[0009] This task has been solved by a method to monitor cell cultures in a medium in a vessel via a computer with a software, where the vessel is connected to at least one sensor measuring the cell cultures, a data model is created and ran by the software using spectral data measured by the at least one sensor to calculate specific process parameters of the cell cultures, wherein the data model is first built and trained by measuring synthetic, cell-free media samples with the at least one sensor including a design space of different ranges of combined media compounds and / or metabolites and / or products and / or process parameters before using the trained data model by the software to perform real-time monitoring of the target cell culture parameters. The idea of using synthetic models is to explore the design space in cell culture media before cultivation of cells. The principle is to reproduce, synthetically and in a cell-free environment, the design space of the parameters of interest which we would like to monitor during the cell culture without having to perform several cell culture runs to create a model. Ranges of the various components to be modeled are performed in cell culture media to be used in a cell culture. It also includes design of experiment to map out the relevant design space resulting in combining various components in the media. Raman spectra are acquired for every condition and a ready to use model can be built for the users who can run their first cell culture with the Raman probe which provides quantitative data in-line and in real-time. Using synthetic samples reduces drastically the time of data acquisition to have a first model, in this case, the data generation take only few days or a week. It also allows cost reduction and transposability to other cell culture media, cell lines, potentially volume scale.

[0010] Preferred use of the process include, for example, but are not limited to that:

[0011] • the relevant design space of different ranges of the combined various media components is mapped out by using a design of experiment approach.

[0012] • as at least one Raman spectroscopy probe is used providing quantitative Raman real-time data for every relevant media condition.

[0013] • a bioreactor is used as vessel, or a beaker is used as a vessel or any container suitable for cell culture is used as a vessel

[0014] • or any liquid container in which a cell culture fluid could be measured is used as a vessel.

[0015] • the data model building is done by measuring the relevant design space either in-line in a bioreactor or a vessel or at-line using the design of experiment approach in a vessel or both.

[0016] • for the measuring in a bioreactor the media are prepared without, or with low concentration of, the component to be monitored, sterilized and put into a sterile bioreactor with the Raman probe, wherein the media is under agitation, aeration and temperature control, using the same set point as a typical bioprocess with a cell culture wherein the metabolite of interest can be added to create a range of concentrations.

[0017] • for the measuring in vessel using the design of experiment approach a mix of various media compounds and / or metabolites and / or products and / or process parameters to be monitored, like glucose and lactate, in the media is created to map out the mix of concentrations met in a real bioprocess with a cell culture, wherein the media samples are prepared by adding a solution of glucose and lactate in different concentrations for each media sample mix. The products could be for instance bio- or cell products such as proteins, antibodies, mRNA etc.

[0018] • the model building process is done under cell-free conditions.

[0019] Another solution to this task is an automated system for monitoring cell cultures in a medium comprising a vessel with at least one Raman sensor to measure the cell cultures, a computer being connected to the at least one sensor and a software performed on the computer with a data interface managing the connection to the at least one sensor and providing a data model, being arranged to perform the previously described methods.

[0020] Preferred further developments of the automated system include, for example, but are not limited to that:

[0021] • the at least one sensor is a probe for Raman spectroscopy.

[0022] • the vessel is a bioreactor or a beaker, or any container suitable for cell culture or any liquid container in which a cell culture fluid could be measured.

[0023] • as software an off-line desktop software for data acquisition and / or data modeling, such as SIMCA, Matlab, is used.

[0024] The inventive approach provides several advantages compared to the known prior art. Those include: • A faster model building to achieve quickly a first robust monitoring with less time needed to build a model.

[0025] • The data model is not dependent from a given process, media, scale or cell line.

[0026] • A better correlation and robustness of the model for a given component.

[0027] • A reduction of operation time, consumables, resources, equipment immobilization.

[0028] Detailed description of the invention

[0029] The method and the automated system 1 including the software 5 according to the invention and functionally advantageous developments of those are described in more detail below with reference to the associated drawings using at least one preferred exemplary embodiment. In the drawings, elements that correspond to one another are provided with the same reference numerals.

[0030] The drawings show:

[0031] Figure 1 : a schematic overview about the used Raman spectroscopy system

[0032] Figure 2: a general schematic about the workflow for building and applying the data model

[0033] Figure 3: the measurement set-up for ranges and DoE (Design of Experiment)

[0034] Figure 4: a map showing concentrations of glucose and lactate encountered in a cell culture process, the composition of the DoE samples and the ranges of glucose and lactate are displayed Figure 5: a map showing concentrations of glucose and lactate encountered in a cell culture process and the composition of the DoE samples are displayed

[0035] Figure 6: the table results for glucose and lactate showing on the left side a calibration set (synthetic data) and on the right side a validation set (cell-culture data)

[0036] Figure 7: the respective prediction results for glucose from the built data model compared the measured reference values

[0037] Figure 8: the respective prediction results for lactate from the built data model compared the measured reference values

[0038] Figure 9: a comparison between the results of a classical cell-culture model building known from the prior art and the inventive synthetic model approach for glucose parameter

[0039] Figure 10a: a schematic comparison of the approach of the model building and the monitoring process with the state of the art approach

[0040] Figure 10b: a schematic comparison of the approach of the model building and the monitoring process without the synthetic approach

[0041] Figure 1 shows an example of a used Raman spectroscopy system 1 which is used for the invention. It comprises of the bioreactor 3 itself which contains a biomass with cell cultures, its control unit 2, a Raman probe 6 connected to the bioreactor 3 and a software 5 run by the control unit 2 which uses at least one specific data model 8 to calculate specific cell culture parameters by analyzing real-time spectral data from the Raman probe 6 to the control unit 2. The control unit 2 is preferably a standard computer suitable to control the bioreactor 3. Another option is a microcontroller or a processor integrated in an embedded device with the bioreactor 3. It could also be a standard or industrial personal computer or server or any other suitable device, especially if the local control unit 2 provides the data model 8 itself, because then a higher processing power as usually provided by a microcontroller is required. In another preferred embodiment the data model 8 is provided by a suitable separate computer with the same or another software at a remote location via a data network using a cloud-based service or a memory device.

[0042] Figure 2 shows a general schematic about the workflow for building and applying the data model for its use in the cell culture monitoring process. Two types of measurement approaches are hereby preferably used:

[0043] 1 . The measurement is done in-line of ranges in bioreactor or vessel.

[0044] 2. The measurement is done at-line using a DoE approach in a vessel or a sample tube for example.

[0045] For the model building process, one or both types can be used. Raman spectra from cell free media preparation are acquired to build a data model. The data model hereby described is applied for monitoring a cell culture.

[0046] Figure 3 shows the measurement set-up for ranges and DoE which is performed in a bioreactor or vessel 3. In the following chapters both approaches and the following model building method steps are explained in more detail using respective preferred working examples.

[0047] 1 . Measuring of ranges in a bioreactor 3:

[0048] The cell culture media is prepared without the component to be monitored, sterilized and put into a sterile bioreactor vessel 3 with the Raman probe 6. The media is under agitation, aeration and temperature control as a cell culture would be, using the same or similar set point as a typical cell culture run. Then, a sterile bottle of glucose or lactate is plugged on the bioreactor and connected to a pump. The solution of glucose or lactate is then slowly added in the vessel 3, as a linear ramp and the Raman spectra are collected automatically at a frequent interval. Some samples are taken during the experiment to perform reference measurements for glucose and lactate. The values of glucose and lactate are then determined, using a linear regression for instance, so that each Raman spectra has a matching concentration for the desired molecule (see also Figure 4). When the desired range of the given metabolite is reached, the pump is stopped, and the data collected in order to build the model. The desired ranged is determined depending on the process aimed to be monitored in cell culture with the aim that the synthetic range is larger than the process range: for example, if the range of glucose in cell culture is meant to be 2,5-8 g / L, the range for the model 8 can be 0-10 g / L. The pump rate and the frequency of the Raman acquisition are selected so that a large number of data point can be collected. Here, more than 100 points are collected for each metabolite, running fully automatically during one or several days, meaning there is no need of human intervention for the measurement. One run for each metabolite can be performed to create a full dataset.

[0049] 2. Measuring using DoE in a vessel 3:

[0050] The DoE approach aims to create mix of the different metabolites, such as glucose and lactate, in cell culture media to map out the mix of concentrations met in a real process. Samples are prepared in cell culture media without the component to be monitored, solution of glucose and lactate are added in different concentrations for each sample mix. A DoE is built to choose the composition of each mix across the desired range. The desired range is determined depending on the process aimed to be monitored in cell culture with the aim that the synthetic sample mix is larger than the process range. For example, the concentrations of glucose and lactate encountered in a cell culture process are displayed as a map in Figure 5. The composition of the synthetic sample mix is chosen to map out the concentration mix displayed in Figure 5. There is a plot of cell cultures data point (ZN1 -6), one cross represents one sample of the cell culture run regarding its concentration of glucose and lactate is shown. The dots represented the concentration chosen for the synthetic sample mix. In this example, 13 sample mixes were chosen to map out the design space, each 13 samples mix were prepared in triplicate and a Raman measurement was taken with the Raman probe 6 , resulting in 13x3 Raman spectra. The reference values for each sample are measured in order to build the respective data models.

[0051] The data acquired with the Raman probe 6 are preprocessed using a preprocessing method and exported, as for example in the following order:

[0052] 1 ) SNV (normalization) calculated on the water region of the spectra

[0053] 2) Savitzky-Golay first order derivative, second order smoothing polynomial and five points window (corresponding to 15 cm’1)

[0054] 3) Raman Shifts selection as following: 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)

[0055] The exported file, e.g. a .csv file, containing the preprocessed spectral data and the reference values is then imported into a suitable analyzing software module (here used: Bio4C® PAT Chemometric Expert) to build the at least one PLS data model. Once built, the data model is exported as a file which was then uploaded into another software module (here used: Bio4C® PAT Raman Software 5) to perform predictions of the cell culture batches (ZN2- 10 batches).

[0056] The results for glucose are displayed in the upper table of Figure 6, where the left side refers to the calibration set (synthetic data) and the right side refers to the validation set (cell-culture data). The abbreviations have the following meaning:

[0057] LV: number of Latent Variables for the PLS models.

[0058] N: number of calibration samples. Calibration range: range of concentration for the analyte.

[0059] RMSE: Root Mean Squared Error (of Calibration (C), of Cross-Validation

[0060] (CV), of Prediction (P))

[0061] The results for lactate are displayed in the lower table of Figure 6 in a similar manner. Figure 7 and Figure 8 show the respective prediction results for glucose (Figure 7) and lactate (Figure 8) from the built data model 8 compared to the measured reference values.

[0062] Figure 9 shows a comparison between the results of a classical cell culture model 8 building known from the prior art and the inventive synthetic model 8 approach for glucose parameter. For that approach, the ZN-8 batch was used as an example. In the results shown in Figure 9 it can be clearly seen that the built synthetic data model 8 outperforms the real cell culture model used for monitoring by a factor of approx. 3 for glucose parameter. Figure 10a and 10b give additionally a schematic comparison of both approaches. Both Figures show the approach of the model building and the monitoring process. In Figure 10a a process time of three to six months is required. Demonstrated performances are on cell densities, main metabolies, including glucose, lactate and others. The invented approach using the synthetic data model 8 requires only a process time of one to two weeks. Its superior performance has been demonstrated for glucose parameter. For lactate parameter the data is also promising, but its superiority not been determined in numbers yet.

[0063] In the following three further preferred working example will be disclosed. In the first example, also referred as “homemade model” 8, a given user of Raman PAT Platform makes some samples with:

[0064] • Fresh media

[0065] • A range of the component or product to monitor (glucose, lactate, or other)

[0066] • Several components can be simultaneously added in a sample as part of a DoE to account for cross-components interactions

[0067] The Raman spectra are acquired for each sample, the spectra are preprocessed using different algorithms, such as normalization, Savitzky- Golay derivativeand spectral selection. Then, a chemometric data model 8 is built for each component to monitor. The preprocessing used for the models 8 are highly specific and need to be optimized for each component. The data model 8 is then used to measure in real-time the concentration of those components in cell culture using the same media. The data model 8 could be used on different media and different cell lines but performances have to be confirmed.

[0068] A second example is the embedded models which have the same principle as the homemade model 8 described above except that the data model 8 is built in factory and embedded in the system to be delivered to the user. So that the user can directly use the data model 8 without prior model building or data acquisition required from them. In this case, the data model 8 can be built with the same medium that is used on the instrument or can leverage data acquired on different Raman analyzers 1 to be robust enough.

[0069] The third example is a big data library from different Raman analyzers 1 , the software or database being either embedded or cloud based. In this example, Raman spectra are acquired on different Raman analyzers 1 for different media. All data are used to build a generic model that can be used with any media. Each synthetic preparation can be added to the data library to make the model 8 more and more robust. Performances of the model have to be tested to check if the model is efficient enough for the chosen medium. The big data library could be either installed at the used factory and updated during maintenance; or shared through a cloud and updated in real-time by other users.

[0070] The software 5 can automatically select data in the big database to create a respective data model based 8 on the information about the process to monitor, e.g. parameters like the used media, type of process and others.

[0071] List of references

[0072] 1 Raman spectroscopy system

[0073] 2 Computer

[0074] 3 Vessel (Bioreactor)

[0075] 4 User interface

[0076] 5 Software

[0077] 6 Raman Probe

[0078] 7 User

[0079] 8 Data Model

Claims

Patent claims1 . A method to monitor cell cultures in a medium in a vessel (3) via a computer(2) with a software (5), where the vessel (3) is connected to at least one sensor (6) measuring the cell cultures (5), a data model (8) is created and ran by the software (5) using spectral data measured by the at least one sensor (6) to calculate specific process parameters of the cell cultures, wherein the data model (8) is first built and trained by measuring synthetic, cell-free media samples with the at least one sensor (6) including a design space of different ranges of combined media compounds and / or metabolites and / or products and / or process parameters before using the trained data model (8) by the software (5) to perform real-time monitoring of the target cell culture parameters.

2. A method according to claim 1 , wherein the relevant design space of different ranges of the combined various media compounds and / or metabolites and / or products and / or process parameters is mapped out by using a design of experiment approach.

3. A method according to one of the previous claims, wherein the at least one sensor (6) is a Raman probe and is used to provide real time data for every relevant bioprocess condition.

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

5. Method according to one of the previous claims, wherein the data model (8) building is done by measuring the relevant design space either in-line in a bioreactor or a vessel (3) or at-line using the design of experiment approach in a vessel (3) or both.

6. Method according to claim 5, wherein for the measuring in a bioreactor (3) the media are prepared without, or with a low concentration of, the component which is to be monitored.

7. Method according to claim 6, wherein the media is sterilized and put into a sterile bioreactor with the Raman probe (6), wherein the media is under agitation, aeration and temperature control, using the same or similar set point as a typical bioprocess with a cell culture.

8. Method according to claim 5, wherein for the measuring in the vessel (3) using the design of experiment approach a mix of various media compounds and / or metabolites and / or products and / or process parameters to be monitored, like glucose and lactate, in the media is created to map out the mix of concentrations met in a real bioprocess with a cell culture, wherein the media samples are prepared by adding a solution of glucose and lactate in different concentrations for each media sample mix.

9. Method according to one of the previous claims, wherein the model building process is done under cell-free conditions.

10. Automated system for monitoring cell cultures in a medium comprising a vessel with at least one sensor (6) to measure the cell cultures process parameters, a computer (2) being connected to the at least one sensor (6) and a software (5) performed on the computer (2) with a data interfacemanaging the connection to the at least one sensor (6) and providing a data model (8), being arranged to perform one of the previous claims.11 .Automated system according to claim 10, wherein the at least one sensor (6) is a probe for Raman spectroscopy.

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

13. “Automated system according to any of claims 10 to 12, wherein as software an off-line desktop software for data acquisition and / or data modeling, such as SIMCA, Matlab is used.

Images