METHOD FOR CONTROLLING CONTINUOUS CHROMATOGRAPHY AND MULTI-COLUMN CHROMATOGRAPHY ARRANGEMENT

Abstract

Methods of controlling at least one multi-column chromatography arrangement for the continuous process of purifying biopharmaceuticals are provided. The methods include introducing a first multi-component mixture into a column of the multi-column chromatography arrangement; detecting at least one multivariate signal by at least one detector; calculating at least one process parameter on the basis of the multivariate signal by at least one data-processing program of a computing unit via application of a chemometric method; and controlling the purification process via control of at least one controllable control element on the basis of the at least one process parameter.

Claims

1. A method for controlling at least one multi-column chromatography arrangement for a continuous process of purification of biopharmaceuticals, comprising: introducing a first multi-component mixture into a column of the multi-column chromatography arrangement, detecting at least one multivariate signal by at least one detector, calculating at least one process parameter based on the multivariate signal by at least one data processing program of a computing unit by applying a chemometric method and controlling the purification process by controlling at least one controllable control element based on the at least one process parameter.

2. The method according to claim 1, wherein the method comprises the further steps of: comparing the at least one process parameter with at least one reference parameter and determining at least one control signal on the basis of the process parameter and wherein the purification process is controlled by controlling the control signal.

3. The method according to claim 1, wherein the at least one data processing program calculates the concentration of at least one component of the at least one multi-component mixture as process parameter.

4. The method according to claim 1, wherein the at least one data processing program calculates at least one of the following process parameters: pH value, conductivity, absorption of the effluent, target protein content, concentration of coeluting contaminants, product concentration, purity, yield, production rate and in and out of specification.

5. The method according to claim 1, wherein the at least one detector is configured to record at least one multivariate signal, wherein the multivariate signal comprises one or more of the following signals: UV spectroscopy signal, vis spectroscopy signal, fluorescence spectroscopy signal, scattered light signal, infrared spectroscopy signal and Raman spectroscopy signal.

6. The method according to claim 1, wherein the calculation of at least one process parameter by means of the one data processing program is carried out by means of at least one of the following chemometric methods: Partial Least Squares Regression and/or calculations by means of a neural network.

7. The method according to claim 1, wherein the method further comprises: determining the product concentration in the multi-component mixture at the feed device by means of the detector, which is preferably arranged at an output of a column, calculating the current product mass loaded onto a column by means of the determined product concentration and a current flow rate, comparing the currently loaded product mass with a control value, and controlling at least one controllable control element based on the currently loaded mass for controlling the purification process.

8. The method according to claim 1, wherein the method further comprises: evaluating a multivariate signal of the at least one detector by means of the at least one computing unit, wherein the detector is preferably arranged at an output of a column so that the component concentration in the multi-component mixture is detected at the discharge device, determining the saturation and/or breakthrough point of the at least one column by means of the detected component concentration and controlling the at least one control element in order to set the interconnection of the columns based on the detected saturation and/or the breakthrough point.

9. The method according to claim 1, wherein the method further comprises: detecting the concentration of one or several, in particular all single components by evaluating the at least one multivariate signal of at least one detector by means of the at least one computing unit, wherein the at least one detector is arranged between two columns, calculating a mass percent of at least one component, comparing the determined mass percent with a control value by means of the computing unit and controlling the at least one control element in order to set the interconnection of the columns and/or fractioning of a product and/or rejecting of fractions based on the control value.

10. The method according to claim 1, wherein the method further comprises the controlling of a recycling of incompletely separated areas.

11. The method according to claim 1, wherein the method further comprises: calculating new optimized process parameters by means of a mathematical model on the basis of the at least one calculated process parameter; and setting the control element on the basis of the calculated process parameters in order to obtain new optimized process parameters.

12. A multi-column chromatography arrangement usable for a continuous process of purification of biopharmaceuticals, comprising: at least one first separation column and one second separation column, each with at least one input and each with at least one output, at least one first feed device suitable for transporting at least one first multi-component mixture, wherein the at least one first feed device is connected to the at least one input of the first separation column, at least one first discharge device suitable for discharging a second multi-component mixture, wherein the at least first discharge device is connected to the at least one output of the first separation column, at least one second feed device suitable for transporting a second multi-component mixture, wherein the at least second feed device is connected to at least one input of a second separation column, at least one second discharge device suitable for discharging a third multi-component mixture, wherein the at least second discharge device is connected to the at least one output of the second separation column, at least one detector for detecting a multivariate signal, and at least one computing unit with at least one data processing program with at least one chemometric calculation method, wherein the at least one first discharge device is connected to the at least one second feed device; wherein at least one of the first and second feed devices and/or at least one of the first and second discharge devices have at least one controllable control element; wherein the at least one detector is arranged before at least one input and/or after at least one output; wherein the at least one detector is coupled to the at least one computing unit; wherein the at least one computing unit is coupled to the at least one controllable control element and wherein the at least one computing unit is configured to: calculate the at least one process parameter based on the multivariate signal, and control at least one controllable control element based on the at least one process parameter.

13. The multi-column chromatography arrangement according to claim 12, wherein the computing unit is further configured to: compare at least one process parameter with at least one reference parameter; determine at least one control signal based on the process parameter and control the purification process by means of controlling the control signal.

14. The multi-column chromatography arrangement according to claim 12, in a continuous process of purification of in particular biopharmaceuticals, wherein the multi-component mixtures are complex mixtures, in particular of cell culture supernatants and/or supernatants of fermentations.

15. The multi-column chromatography arrangement according to claim 13, in a continuous process of purification of in particular biopharmaceuticals, wherein the multi-component mixtures are complex mixtures, in particular of cell culture supernatants and/or supernatants of fermentations.

Description

[0115] In the following, the subject matter of the invention is described by means of embodiments.

[0116] FIG. 1 shows an experimental set-up in batch mode of a chromatography arrangement (schematic representation of Akta pure of the company GE Healthcare, Chalfont St. Giles, UK) with a diode array detector (DAD).

[0117] FIG. 2 shows an exemplary flow of information for the calibration of a mathematical model.

[0118] FIG. 3 shows an exemplary statistical experimental plan for model calibration and validation for the control of the load duration.

[0119] FIG. 4 shows an exemplary comparison of a PLS model prediction with a reference analysis.

[0120] FIG. 5 shows an exemplary comparison of an NN model prediction with the reference analysis.

[0121] FIG. 6 shows an exemplary comparison of a PLS model prediction with the reference analysis.

[0122] FIG. 7 shows an exemplary comparison of the NN model prediction with the reference analysis.

[0123] FIG. 8 shows a pipeline and instrument flow chart for controlling continuous chromatography.

[0124] FIG. 9 shows a flow diagram for general preparatory works.

[0125] FIG. 10 shows a flow diagram for a general example for a process control.

[0126] FIG. 11 shows a flow diagram of a specific example for a preparatory work.

[0127] FIG. 12 shows a flow diagram for a specific example for a process control.

[0128] FIG. 13 shows two absorption spectra of antibody and antibody aggregates.

[0129] FIG. 14 shows the spectral difference in the UV area between antibody and antibody aggregates (A) and an associated difference spectrum (B).

[0130] FIG. 1 shows an experimental set-up for an experiment with a chromatography arrangement in batch mode. A stock of a mobile phase 6, a control element 4, a column 2, a feed device 8, internal detectors 10, a diode array detector (DAD) 12 and a fraction device 14 are shown. The chromatography arrangement exemplarily described here is the chromatography arrangement distributed under the name Akta pure by GE Healthcare, Chalfont St. Giles, UK. Additionally, a DAD 12 is arranged in the chromatography arrangement, which is distributed under the name Ultimate 3000 by Dionex, Fisher Scientific, USA. As described above, several chromatographic separations can be performed under different conditions for training and validation of the mathematical models used. In this, the effluent of column 2 is guided via the DAD 12 and subsequently fractioned in all experiments. In this, DAD 12 continuously taps absorption spectra.

[0131] As shown in FIG. 2, fractions 20 can then be examined with an external reference analysis 18 (analytical chromatography) with regard to the concentration of relevant species. Absorption spectra 16 have been averaged corresponding to the time duration of a fraction 20. Subsequently, the mathematical models are calibrated with the training data and verified on the basis of the validation data.

[0132] A first embodiment shows the control of the load duration. The training and validation of the mathematical model take place on the basis of a statistical experimental design, shown in FIG. 3. In this, the feed is added variable percentages of product and contaminants in order to replicate process variation during a continuous process. Additionally, the retention time of the product in column 2 during the load phase is varied. The center point of the statistical experimental plan represents the validation batch.

[0133] From the data both a Partial Least Squares (PLS) Regression model and an Artificial Neural Network (NN) are calibrated or trained, respectively. A comparison of the model prediction and the reference analysis for the training and validation data is shown in FIG. 4 for the PLS model and in FIG. 5 for the NN model.

[0134] The calibrated PLS model can subsequently be used for controlling the load phase during a protein chromatography. In this, it is advantageous that the calibrated PLS model can be used for the online process monitoring and thus allows for a continuous purification of multi-component mixtures without an offline analysis of the feed being required. Since the product concentration is variable at the same time and also the capacity of the column can change by ageing, in every new batch the titer must first be determined offline before the process can be continued.

[0135] Additionally, extensive studies must be carried out in order to examine the column ageing. In contrast, by means of the deconvoluted multivariate signal it is possible to determine at the output of the column if the product breaks through. By means of this information a corresponding valve can be connected as control element and the pump operation of the feed be terminated at a suitable point of time.

[0136] Another exemplary embodiment is the control of the fractioning of a target product.

[0137] The training and validation of the mathematical model take place by means of several chromatography runs with variable length of the salinity gradients (1, 3, 5, 7 column volume [CV]) in a cation exchange step. In this, the variation of the gradient length results in different concentration ratios of the species in the effluent which span the calibration space. Additionally, within this space a validation batch can be carried out.

[0138] From the data, both a Partial Least Squares Regression (PLS) model and an Artificial Neural Network (NN) are calibrated. A comparison of the model prediction and the reference analysis for the training and validation data is shown in FIG. 6 for the PLS model and in FIG. 7 for the NN.

[0139] The calibrated PLS model is subsequently used for controlling and thus optimizing the fractioning of a target protein in real-time in a polishing step. The problem solved this way is that the separation of product and contaminants may vary from batch to batch due to process variations. Since product and contaminants cannot be selectively detected this leads to variable product qualities. In contrast, by means of the deconvoluted multivariate signal it is possible to selectively detect at which point of time the desired product quality is reached, and by means of this information the valve for fractioning the product can be controlled.

[0140] In total, the method according to the invention for controlling and monitoring processes can be used to control chromatographic separations. This is in particular essential for continuous processes, since in them the process cannot be stopped for an offline analysis and errors may accumulate from cycle to cycle. However, by means of the method according to the invention the stationary state can easily be controlled.

[0141] In the following, the execution of the embodiments in continuous chromatography is exemplarily described.

[0142] The above described method for process control can be used in continuous chromatography. In this, the detected signals and evaluation by means of mathematical methods remains identical to the batch mode. However, what is different is how the deconvoluted signal is used for controlling the valves. Such a concept is shown in FIG. 8. Depending on the application, detectors are integrated in different places into the pipelines 22 of the multi-column chromatography arrangement.

[0143] Exemplarily, the method for controlling the load duration can be used in the continuous process. In the continuous production of biopharmaceuticals the product concentration can vary by variability in the upstream and cannot be detected in the feed line with traditional, univariate detectors. However, since columns 26, 28, 30 have a limited capacity, which moreover can change by ageing of the columns, a detection of the loaded product mass is critical. This can be achieved in real time by means of the deconvoluted, multivariate signals of detectors 32 and 34 in FIG. 8.

[0144] Assuming that the feed is pumped via detector 32 onto column 26 and subsequently guided via detector 34 onto column 28. By means of the integration of the deconvoluted multivariate signal of detector 32 the loaded product mass can be continuously calculated.

[0145] Since the capacity of the columns may change by ageing, the controlling of the interconnection of columns takes place by means of the deconvoluted multivariate signal of detector 34. The latter is always located between loaded and receiving column. If a specific threshold value of the product concentration is reached at the output of column 26, the position of valves 36 and 38 can be changed in such a way that the feed is guided onto column 28 and via or past, respectively, detector 34 onto column 30. The above described process is cyclically repeated continuously for all columns 26, 28, 30.

[0146] Another example for an application of the method of the present invention is the control of the fractioning in the continuous process.

[0147] In the product eluation of column 26, 28, 30 in the continuous method the decision must be taken via valve 40 whether the effluent is fractioned, recycled or rejected. By means of process variations the separation of product and contaminants can change in the continuous process.

[0148] Since product and contaminants cannot be selectively detected by means of traditional methods, a process-oriented control of valve 40 via univariate measurement methods can be complex. In this, in the worst case errors may accumulate and the process can become instable. This can negatively affect the product quality and yield. By means of the deconvoluted multivariate signal of detector 42, however, it is possible to selectively detect at which point of time the desired product quality or yield is given as preferred process parameter. Based on this information valve 40 can be controlled.

[0149] FIGS. 9 and 10 show exemplary flow diagrams for preparatory work and for the process control for an exemplary specific case. As the case may be, a calibration of a chemometric model is performed. This does not apply when using a multivariate curve resolution. The calibration first takes place in batch mode as well as using an offline reference analysis.

[0150] Thus it is possible by means of the calibration to define a target function and to determine the associated threshold values, such as a specific purity and/or the maximum mass to be loaded.

[0151] The data thus obtained are then used for the continuous multi-column chromatography for purifying biopharmaceuticals (FIG. 10). At least one multivariate signal in a multi-column chromatography arrangement is recorded. The recording takes place by means of detectors in the feed and/or discharge of the columns.

[0152] The at least one recorded multivariate signal is then evaluated by means of at least one chemometric model or a chemometric method, respectively, for instance in a calculation of the individual concentrations and/or classification into correct or incorrect.

[0153] Based on the results of the chemometric method the process decision is taken, for instance a change of the interconnection of the columns and/or the flow rate and/or the gradient.

[0154] FIG. 11 shows an exemplary flow diagram for the preparatory work for an exemplary general case.

[0155] The calibration of the chemometric model for a complex feed of for instance a bioreactor is first made possible by means of calibration tests in batch mode. A feed of for instance a bioreactor traditionally has a varied product concentration and quality. In the calibration tests in batch mode therefore important parameters are varied, for instance concentration of the product protein, amount and composition of the contaminants. Further, the maximum load mass on a column is set. These variations are set based on preliminary tests, process understanding, etc. Subsequently, the change of the effective interconnection of columns is set.

[0156] FIG. 12 shows an exemplary flow diagram for the process control for an exemplary general case.

[0157] The first column of a multi-column chromatography arrangement is loaded with a complex mixture with variable product concentrations. A multivariate signal is detected in the feed to the column. The signal is then evaluated by means of the trained model and thus for instance the product concentration is determined. As a further step then summing up or integration, respectively, of the masses of the individual components of the mixture takes place.

[0158] The term mass in this case refers to the currently loaded product mass on the first column. Moreover, in the concretely described example only product concentration and loaded product mass are determined.

[0159] The next step is the change to a second column of the multi-column chromatography arrangement if the mass exceeds the predefined control value. The process can then be repeated for the second column.

[0160] FIG. 13 shows two exemplary absorption spectra. One absorption spectrum refers to an antibody and is shown in FIG. 13 as solid line. The spectrum represented in FIG. 13 as dotted line refers to an antibody aggregate. It is clearly discernible how little the differences between the spectra are. In particular, it had been assumed before the invention that due to the slight differences between such spectra such spectra cannot be used for controlling a continuous multi-column chromatography. However, according to the present invention it was recognized that it is possible by means of chemometric methods, in particular PLS and/or neural networks, due to the low but existing spectral differences to control a multi-column chromatography in such a way, in particular to take process decisions, in order to control a purification process in such a way that a biopharmaceutical and/or protein and/or protein mixture and/or otherwise requested product to extract with sufficient purity. Thus for instance a requested antibody can be selectively quantified in or extracted from, respectively, multi-component mixtures, i.e. purified, despite the spectra only differing in the mAU area.

[0161] FIG. 14a shows the UV spectra of two components in a component mixture which are denominated IgG-01 and IgG-02. FIG. 14b shows a difference spectrum of both spectra of FIG. 14a. By means of the method according to the invention it is possible to selectively quantify both antibodies IgG-01 and IgG02 in a mixture and extract, i.e. purify in each case, both from the mixture, though, as discernible from FIG. 14b, the spectral differences are smaller by about three dimensions than the standardized spectrum intensities.