Multivariate Spectroscopic Analysis and Monitoring for Biomanufacturing

Abstract

The present disclosure features a method comprising obtaining a vibrational spectrum of a solution in a biomanufacturing system (804), analyzing the vibrational spectrum using a first chemometric model (806) to determine a first quality attribute value associated with the solution, analyzing the vibrational spectrum using a second chemometric model (806) to determine a second quality attribute value associated with the solution (812), and adjusting at least one parameter of a purification unit of the biomanufacturing system based on at least one of the first quality attribute value and the second quality attribute value (810).

Description

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS

[0002] This application claims priority to the following U.S. provisional applications, each of which is incorporated herein by reference in its entirety: 62 / 637,891, filed March 2, 2018; 62 / 673,845, filed May 18, 2018; and 62 / 729,402, filed September 10, 2018. Technical Field

[0003] The present disclosure relates to systems and methods for use in integrated continuous biomanufacturing systems. Background Art

[0004] Mammalian cells containing nucleic acids encoding recombinant proteins are often used to produce therapeutically or commercially important proteins. Integrated continuous biomanufacturing is an important aspect of reducing the costs associated with such protein-based therapies. Monitoring systems are used in biomanufacturing to evaluate various biological products and process conditions. Summary of the invention

[0005] Integrated continuous biomanufacturing of therapeutic protein substances and other biomolecules offers great promise for the future production of life-saving drugs and for facilitating widespread adoption of therapies that rely on the availability of such biomolecules. Dual-column and multi-column chromatography systems with various configurations are available for industrial-scale biomanufacturing. In such systems, process monitoring of various eluent streams can be used to adjust process-related parameters and control product attributes, for example, selective collection of eluent streams from certain columns and adjustment of solution buffer properties (e.g., pH).

[0006] The present disclosure is characterized in that a method and system for determining solution characteristics using a real-time or near real-time sensor integrated online with a chromatographic system and / or a bioreactor and coupled with an electronic controller (analyzing the information measured by the sensor), the solution characteristics are, for example, solute concentration, charge distribution of the analyte of interest, process and product impurities, molecular integrity, aggregation state, and pH. The infrared spectrum of the solution can be continuously monitored, and a chemometric model can be used to accurately characterize the quantitative chemical, physical, and / or biological properties of various analytes in the solution simultaneously. The spectrum can be obtained online from the flowing solution, thereby measuring with little or no interruption of the manufacturing process. In addition, the chemometric model can extract quantitative analyte information in real time or near real time, allowing rapid feedback and control of parameters and operations related to the biomanufacturing process.

[0007] In a first aspect, the disclosure features a method comprising obtaining a vibrational spectrum of a solution in a biomanufacturing system, analyzing the vibrational spectrum using a first chemometric model to determine a first quality attribute value associated with the solution, analyzing the vibrational spectrum using a second chemometric model to determine a second quality attribute value associated with the solution, and adjusting at least one parameter of a purification unit of the biomanufacturing system based on at least one of the first quality attribute value and the second quality attribute value.

[0008] Embodiments of the method may include any one or more of the following features.

[0009] The method may include using a biomanufacturing system to produce at least one of a protein-based therapeutic substance, a nucleic acid-based drug, and a gene therapy drug. The protein-based therapeutic substance may include at least one of a protein, a peptide, an antibody, and an enzyme. The nucleic acid-based drug may include at least one of a DNA, a plasmid, an oligonucleotide, an aptamer, a DNA enzyme, an RNA aptamer, an RNA bait, a microRNA fragment, and a small interfering RNA fragment.

[0010] For a biological product produced by the biomanufacturing system, the first quality attribute and the second quality attribute can each be independently selected from a product quality attribute, a product-related impurity, and a process-related impurity.

[0011] Obtaining the vibrational spectrum may include directing radiation to be incident on the solution and measuring attenuated totally reflected radiation from the solution. Radiation may be incident on the solution through a radiation window, and the attenuated totally reflected radiation may pass through the radiation window before being measured. The method may include measuring the attenuated totally reflected radiation from the solution as the solution flows relative to the radiation window.

[0012] The radiation window may form part of the flow cell.The incident radiation may comprise infrared radiation.

[0013] The first chemometric model may include a first set of principal vibrational components associated with a first mass attribute. The first chemometric model may include at least three principal vibrational components (e.g., at least five principal vibrational components). Analyzing the vibrational spectrum using the first chemometric model may include calculating the first mass attribute value based on the first set of principal vibrational components. Calculating the first mass attribute value may include determining the value as a linear function of the first set of principal vibrational components.

[0014] The second chemometric model may include a second set of principal vibrational components associated with the first mass attribute. The second chemometric model may include at least three principal vibrational components (e.g., at least five principal vibrational components). The first set of principal vibrational components and the second set of principal vibrational components may have no common members. Analyzing the vibrational spectrum using the second chemometric model may include calculating the second mass attribute value based on the second set of principal vibrational components. Calculating the second mass attribute value may include determining the value as a linear function of the second set of principal vibrational components.

[0015] The solution may include a solution discharged from a purification unit of the biomanufacturing system. The method may include purifying the solution in the purification unit and then obtaining a vibrational spectrum of the solution. The purification unit may include a chromatographic column.

[0016] The method may include obtaining the vibrational spectrum by measuring radiation from the solution as the solution flows between a first purification unit and a second purification unit of the biomanufacturing system.The first purification unit and the second purification unit may each include a chromatography column.

[0017] The method may include obtaining the vibrational spectrum by measuring radiation from the solution after the solution exits a final purification unit of the biomanufacturing system.

[0018] The solution may be a first solution, and the method may include: obtaining a vibrational spectrum of a second solution in a biomanufacturing system, analyzing the vibrational spectrum of the second solution using a first chemometric model to determine a first quality attribute value of the second solution, and analyzing the vibrational spectrum of the second solution using a second chemometric model to determine a second quality attribute value of the second solution. The first solution may flow between a first purification unit and a second purification unit in the biomanufacturing system, and the second solution may flow between the second purification unit and a third purification unit in the biomanufacturing system. The method may include adjusting the at least one parameter based on at least one of the first quality attribute value and the second quality attribute value of the first solution and the first quality attribute value and the second quality attribute value of the second solution.

[0019] The method may include obtaining a vibrational spectrum and determining a first quality attribute value and a second quality attribute value within a time period of 30 seconds or less from the time the radiation is incident on the solution. The time period may be 10 seconds or less (e.g., 2 seconds or less).

[0020] The method may include repeating a plurality of steps of obtaining a vibrational spectrum, analyzing the vibrational spectrum using a first chemometric model, and analyzing the vibrational spectrum using a second chemometric model to determine a time series of first and second quality attribute values ​​associated with successive portions of the solution. The method may include adjusting at least one parameter based on at least one of the time series of the first and second quality attribute values.

[0021] The method may include obtaining a set of one or more calibration spectra representative of the solution, and analyzing the set of calibration spectra to determine the first set of principal vibrational components and the second set of principal vibrational components. The first chemometric model may include a first set of coefficients associated with the first set of principal vibrational components, and the second chemometric model may include a second set of coefficients associated with the second set of principal vibrational components, and the method may include determining the first set of coefficients and the second set of coefficients based on a regression analysis.

[0022] The method may include analyzing the vibrational spectrum using a third chemometric model to determine a third quality attribute value associated with the solution. The third chemometric model may include a third set of primary vibrational components associated with the third quality attribute. The third chemometric model may include at least three primary vibrational components (e.g., at least five primary vibrational components). Analyzing the vibrational spectrum using the third chemometric model may include calculating the third quality attribute value based on the third set of primary vibrational components. Calculating the third quality attribute value may include determining the value as a linear function of the third set of primary vibrational components. The method may include adjusting at least one parameter based on at least one of the first quality attribute value, the second quality attribute value, and the third quality attribute value.

[0023] The method may include analyzing the vibrational spectrum using a fourth chemometric model to determine a fourth quality attribute value associated with the solution. The fourth chemometric model may include a fourth set of primary vibrational components associated with the fourth quality attribute. The fourth chemometric model may include at least three primary vibrational components (e.g., at least five primary vibrational components). Analyzing the vibrational spectrum using the fourth chemometric model may include calculating a fourth quality attribute value based on the fourth set of primary vibrational components. Calculating the fourth quality attribute value may include determining the value as a linear function of the fourth set of primary vibrational components. The method may include adjusting at least one parameter based on at least one of the first quality attribute value, the second quality attribute value, the third quality attribute value, and the fourth quality attribute value.

[0024] At least one of the first quality attribute and the second quality attribute may be selected from concentration, aggregation, charge variant distribution, purity, glycan profile, identity, and integrity. At least one of the first quality attribute and the second quality attribute may be selected from protein fragments, nucleic acid fragments, nucleic acid variants, empty capsids, and vector impurities. At least one of the first quality attribute and the second quality attribute may be selected from host cell proteins, residual host DNA, residual column ligands, impurity concentration, impurity amount, residual helper virus, residual helper virus protein, and residual helper virus DNA.

[0025] Unless expressly stated otherwise, embodiments of the method may also include any other features disclosed herein, including combinations of features disclosed individually in connection with different embodiments.

[0026] On the other hand, the present disclosure features a biomanufacturing system, which includes: a bioreactor configured to produce a solution containing a biological product; a purification unit configured to receive the solution; a radiation source configured to direct radiation incident on the solution; a detection device configured to measure radiation from the solution; and a system controller connected to the bioreactor and the detection device and configured to: receive a measurement signal from the detection device, the measurement signal corresponding to information about a vibrational spectrum of the solution; analyze the information using a first chemometric model to determine a first quality attribute value associated with the solution; analyze the information using a second chemometric model to determine a second quality attribute value associated with the solution; and adjust at least one parameter of the purification unit based on at least one of the first quality attribute value and the second quality attribute value.

[0027] Embodiments of the system may include any one or more of the following features.

[0028] The system may include a flow cell positioned to pass the solution through the flow cell, and the radiation source directs the radiation to be incident on the solution while the solution is in the flow cell.

[0029] The biological product may include at least one of a protein-based therapeutic substance, a nucleic acid-based drug, and a gene therapy drug. The protein-based therapeutic substance may include at least one of a protein, a peptide, an antibody, and an enzyme. The nucleic acid-based drug may include at least one of a DNA, a plasmid, an oligonucleotide, an aptamer, a DNA enzyme, an RNA aptamer, an RNA bait, a microRNA fragment, and a small interfering RNA fragment.

[0030] For a biological product produced by the biomanufacturing system, the first quality attribute and the second quality attribute can each be independently selected from a product quality attribute, a product-related impurity, and a process-related impurity.

[0031] The detector may include a total internal reflection sensor configured to measure attenuated total reflected radiation from the solution. The total internal reflection sensor may be integrated with a portion of the flow cell. The controller and detection device may be configured to measure attenuated total reflected radiation from the solution as the solution flows within the flow cell. The incident radiation may include infrared radiation.

[0032] The first chemometric model may include a first set of principal vibrational components associated with the first mass attribute.The first chemometric model may include at least three principal vibrational components (eg, at least five principal vibrational components).

[0033] The controller may be configured to analyze the vibrational spectrum using the first chemometric model by calculating the first mass property value based on the first set of principal vibrational components. The controller may be configured to calculate the first mass property value as a linear function of the first set of principal vibrational components.

[0034] The second chemometric model may include a second set of principal vibrational components associated with the first mass attribute. The second chemometric model may include at least three principal vibrational components (e.g., at least five principal vibrational components). The first set of principal vibrational components and the second set of principal vibrational components may have no common members. The controller may be configured to analyze the vibrational spectrum using the second chemometric model by calculating the second mass attribute value based on the second set of principal vibrational components. The controller may be configured to calculate the second mass attribute value as a linear function of the second set of principal vibrational components.

[0035] The solution may include a solution discharged from the purification unit. The controller may be connected to the purification unit and configured to purify the solution in the purification unit and then obtain information about the vibration spectrum of the solution.

[0036] The purification unit may include a chromatographic column. The purification unit may be a first purification unit of the system, and the system may include a second purification unit configured to receive the solution, wherein the detection device is positioned to measure radiation from the solution as the solution flows between the first purification unit and the second purification unit. The first purification unit and the second purification unit may each include a chromatographic column.

[0037] The purification unit may be the final purification unit of the system.

[0038] The solution may be a first solution, and the detection device may be configured to measure radiation from a second solution, and the controller may be configured to: receive a measurement signal from the detection device, the measurement signal corresponding to information about a vibrational spectrum of the second solution; analyze the information about the second solution using the first chemometric model to determine a first quality property value of the second solution; and analyze the information about the second solution using the second chemometric model to determine a second quality property value of the second solution.

[0039] The purification unit may be a first purification unit and the system may include a second purification unit and a third purification unit, wherein the first solution may flow between the first purification unit and the second purification unit, and the second solution may flow between the second purification unit and the third purification unit. The controller may be configured to adjust the at least one parameter based on at least one of the first quality attribute value and the second quality attribute value of the first solution and the first quality attribute value and the second quality attribute value of the second solution.

[0040] The detection device and the controller can be configured to measure radiation from the solution and determine the first quality attribute value and the second quality attribute value within a time period of 30 seconds or less from the time the radiation is incident on the solution. The time period can be 10 seconds or less (e.g., 2 seconds or less).

[0041] The detection device and the controller may be configured to repeat a plurality of steps of measuring radiation from the solution, analyzing information about a vibrational spectrum of the solution using a first chemometric model, and analyzing the vibrational spectrum using a second chemometric model to determine a time series of first and second quality attribute values ​​associated with successive portions of the solution. The controller may be configured to adjust at least one parameter based on at least one of the time series of the first and second quality attribute values.

[0042] The controller may be configured to obtain a set of one or more calibration spectra representative of the solution, and analyze the set of calibration spectra to determine the first set of principal vibrational components and the second set of principal vibrational components. The first chemometric model may include a first set of coefficients associated with the first set of principal vibrational components, and the second chemometric model may include a second set of coefficients associated with the second set of principal vibrational components, and the controller may be configured to determine the first set of coefficients and the second set of coefficients by performing a regression analysis.

[0043] The controller may be configured to analyze information about the vibration spectrum of the solution using a third chemometric model to determine a third quality attribute value associated with the solution. The third chemometric model may include a third group of main vibrational components associated with the third quality attribute. The third chemometric model may include at least three main vibrational components (e.g., at least five main vibrational components). The controller may be configured to analyze information about the vibration spectrum using the third chemometric model by calculating a third quality attribute value based on the third group of main vibrational components. The controller may be configured to calculate the third quality attribute value by determining the third quality attribute value as a linear function of the third group of main vibrational components. The controller may be configured to adjust at least one parameter based on at least one of the first quality attribute value, the second quality attribute value, and the third quality attribute value.

[0044] The controller may be configured to analyze the information about the vibrational spectrum using a fourth chemometric model to determine a fourth quality attribute value associated with the solution. The fourth chemometric model may include a fourth group of major vibrational components associated with the fourth quality attribute. The fourth chemometric model may include at least three major vibrational components (e.g., at least five major vibrational components). The controller may be configured to analyze the information about the vibrational spectrum using the fourth chemometric model by calculating a fourth quality attribute value based on the fourth group of major vibrational components. The controller may be configured to calculate the fourth quality attribute value by determining the fourth quality attribute value as a linear function of the fourth group of major vibrational components. The controller may be configured to adjust at least one parameter based on at least one of the first quality attribute value, the second quality attribute value, the third quality attribute value, and the fourth quality attribute value.

[0045] At least one of the first quality attribute and the second quality attribute may be selected from concentration, aggregation, charge variant distribution, purity, glycan profile, identity, and integrity. At least one of the first quality attribute and the second quality attribute may be selected from protein fragments, nucleic acid fragments, nucleic acid variants, empty capsids, and vector impurities. At least one of the first quality attribute and the second quality attribute may be selected from host cell proteins, residual host DNA, residual column ligands, impurity concentration, impurity amount, residual helper virus, residual helper virus protein, and residual helper virus DNA.

[0046] Unless explicitly stated otherwise, embodiments of the system may also include any other features disclosed herein, including combinations of features disclosed separately in connection with different embodiments.

[0047] As used herein, "real time" refers to a measurement or process that occurs with a relatively small delay or recurrence period. For example, a "real time" measurement is one in which the total elapsed time interval between the start of measuring spectral information and the time at which a parameter value or other quantity is calculated from that information is 1 minute or less. A periodic real time measurement is a cyclic / periodic measurement in which the time interval between consecutive measurements is 1 minute or less.

[0048] As used herein, a "near real-time" measurement is one in which the total elapsed time interval between the start of measuring spectral information and the time at which a parameter value or other quantity is calculated from that information is between 1 minute and 5 minutes. A periodic near real-time measurement is a cyclic / periodic measurement in which the time interval between consecutive measurements is between 1 minute and 5 minutes.

[0049] As used herein, the term "quality attribute" refers to a parameter value that is used to evaluate the operating conditions of a biomanufacturing system, the integrity of a process performed in such a system, and / or a product derived from such a process. Quality attributes can be product quality attributes that are related to the purity, integrity, yield, and other characteristics of the product produced by the biomanufacturing system; they can be product-related impurity parameters that provide information about the product produced in the system; and they can be process-related purity parameters that provide information about the fidelity of the biomanufacturing process performed in the system.

[0050] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those generally understood by those of ordinary skill in the art to which this disclosure belongs. Although methods and materials similar or equivalent to the methods and materials described herein can be used in the practice or testing of this paper's subject matter, suitable methods and materials are described below. All publications, patent applications, patents, and other references mentioned herein are incorporated by reference in their entirety. In the event of a conflict, the present specification including the definitions shall prevail. In addition, materials, methods, and embodiments are merely illustrative and are not intended to be limiting.

[0051] The details of one or more embodiments are set forth in the accompanying drawings and the description below. Other features and advantages will become apparent from the description, drawings, and claims. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] Figure 1 is a schematic diagram showing an example of a system for measuring infrared spectral information of a process solution in a biomanufacturing system.

[0053] Figure 2 is a schematic cross-sectional view showing an example of an optical fiber.

[0054] Figure 3is a schematic diagram showing an example of a prism that forms an attenuated total reflection interface.

[0055] Figure 4 is a schematic diagram showing an example of a flow cell positioned between two fluid conduits.

[0056] Figure 5 is a schematic diagram showing another example of a flow cell positioned between two fluid conduits.

[0057] Figure 6 is a schematic diagram showing incident radiation rays being refracted at the surface of a prism.

[0058] Figure 7 is a schematic diagram showing an example of a measurement system for measuring infrared spectral information of a fluid in a biomanufacturing system.

[0059] Figure 8 is a flow chart showing example steps that may be performed to analyze vibrational spectroscopy information of a process fluid.

[0060] Fig. 9 is a flow chart showing example steps that may be performed to build and validate a chemometric model for a process fluid property based on vibrational spectroscopy information measured for the process fluid.

[0061] Fig.10 is a schematic diagram showing an example of a biomanufacturing system.

[0062] Fig.11 is a schematic diagram showing an example of a three-column switching technique in a periodic countercurrent chromatography system.

[0063] Fig.12 It is a graph showing the Fourier transform infrared (FTIR) spectrum of a protein A (Protein A) purified antibody sample.

[0064] Fig.13 is a graph showing predicted antibody concentration values ​​calculated from a partial least squares (PLS) model for antibody concentration and the corresponding measured antibody concentration values.

[0065] Fig.14A The figure shows an example of the Fourier self-deconvolution FTIR infrared spectrum of a purified sample of Protein A.

[0066] Fig. 14B is a graph showing the PLS calibration model developed for the aggregation values ​​of the samples and the measured aggregation values ​​of the samples.

[0067] Fig.15A is a graph showing a set of spectra used to construct a PLS model for the host cell protein (HCP) content of a sample.

[0068] Fig. 15B Is displayed for Fig.15A Plot of the PLS model for the HCP content of samples determined from the spectra in Figure 2 and the measured HCP values ​​of the samples.

[0069] Figures 16A-16C are graphs showing three different PLS models of charge variant distribution values ​​for samples.

[0070] Fig.17 is a table showing the antibody concentration values ​​of six samples predicted using the partial least squares model for antibody concentration.

[0071] Figures 18A-18D is a graph showing measurements of viable cell density, cell viability, harvest titer, and specific productivity for a perfusion bioreactor.

[0072] Figures 19A-19D is a graph showing measured values ​​of glucose concentration, glutamine concentration, lactate concentration and ammonium ion concentration at different cell density, pH and perfusion conditions for a bioreactor producing a monoclonal antibody product.

[0073] Figures 20A-20D is a graph showing daily measurements of glucose concentration, harvest titer, lactate concentration, and ammonium ion concentration of the bioreactor medium.

[0074] Figures 21A-21F is a graph showing measured values ​​of glucose concentration, glutamine concentration, IgG concentration, lactate concentration, ammonium ion concentration, and osmotic pressure concentration of the bioreactor culture medium determined from infrared spectroscopy information.

[0075] Figures 22A-22D is a graph showing multiple values ​​of glucose concentration, harvest titer, lactate concentration, and ammonium ion concentration of the bioreactor culture medium determined from infrared spectroscopy information.

[0076] Figures 23A-23C is a graph showing multiple values ​​of glucose concentration, lactate concentration, and ammonium ion concentration of the bioreactor culture medium at a later stage determined from infrared spectroscopy information.

[0077] The same symbols in the drawings represent the same elements. DETAILED DESCRIPTION

[0078] introduction

[0079] Industrial-scale biomanufacturing can be performed in a variety of configurations in dual-column and multi-column chromatography systems. In these complex systems, product yield, quality, and scrap rates are a function of a large number of process-related parameters and steps. During the manufacture of therapeutic proteins and other commercially valuable biomolecules, product outcomes can be strongly affected by these parameters and steps. Therefore, proper control of such parameters and steps is an important aspect of large-scale manufacturing. For example, features and aspects of biomanufacturing systems are disclosed in PCT Patent Application Publication No. WO2014 / 137903, the entire contents of which are incorporated herein by reference.

[0080] The application of appropriate control (including automatic control) of biomanufacturing parameters is facilitated by in-process on-line monitoring of intermediate solution streams, particularly monitoring the concentrations of intermediates and products in such streams and other characteristics of such streams (e.g., pH). Conventional methods for solution monitoring include techniques such as UV absorbance measurement. Unfortunately, however, such methods drift over measurement periods of several days due to factors such as temperature, humidity, ambient light intensity, and local sample inhomogeneities.

[0081] Furthermore, such methods generally do not allow for multiple quantities to be calculated or otherwise determined in real time or near real time from a single measurement (e.g., measurement of an absorbance spectrum). Instead, there is generally a 1:1 correspondence between the measurements made and the quantities determined from such measurements. Thus, for example, to measure the concentration of two different species in a solution stream, two different UV absorbance measurements would be recorded, and each measurement would yield a concentration value for one of the species. Due to the measurement and analysis time required to determine such values, it may not be possible to determine the values ​​of multiple quantities in real time or near real time.

[0082] Disclosed herein are methods and systems for determining quantitative information about analytes and properties in process solutions using infrared spectroscopy measurements in combination with multivariate chemometric models. For example, analytes may include product and intermediate components, waste products, residual reagents, and buffer components. Examples of properties may include, but are not limited to, pH levels, salinity levels, protein / peptide aggregation levels, the amount / concentration of biological products (such as proteins, peptides, and nucleic acids), the concentration / amount of process and product impurities, and the charge variant distribution of proteins.

[0083] Infrared spectroscopy measurements can be performed quickly and with high reproducibility, and calibrated chemometric models are used to predict quality attributes in real time or near real time under various biomanufacturing process conditions and for various reagents, products, byproducts, and impurities. The extensive control provided by the chemometric-based analytical methods disclosed herein allows for more efficient biomanufacturing processes, with less waste, higher product yields, and higher product purity.

[0084] Due to the wavelength of infrared radiation, infrared spectroscopy measurements generally provide information about the vibrational properties of analytes (e.g., reaction products and byproducts in process fluids within a biomanufacturing system). Although other measurement modalities such as Raman spectroscopy can also produce important information about such analytes, infrared measurements can provide certain advantages in certain circumstances. Typically, for example, measurements of infrared spectroscopy information can be performed faster (e.g., in about 30 seconds or less) than corresponding Raman measurements (which can take between 10 and 15 minutes). Therefore, infrared measurements can be better suited for real-time and near real-time process monitoring and control applications.

[0085] To date, the manufacturing process has been carried out in aqueous solution, and water is at about 1640 cm -1 Due to strong water absorption at high temperatures, infrared measurements have limited application in upstream and downstream process monitoring in biomanufacturing operations for biopharmaceuticals. Process water inherently interferes with conventional analytical calculations based on infrared measurements, sometimes to the point where quantitative information cannot be determined from such measurements. However, even in the presence of strong water absorption, the chemometric methods disclosed herein can reliably predict quantitative values ​​for various parameters. Thus, the methods disclosed herein enable routine, high-throughput process monitoring and control using infrared spectroscopy measurements.

[0086] Additionally, IR spectroscopy measurements are more sensitive to changes in analytes and process fluid properties than Raman measurements. Combined with the rate at which such measurements can be made, IR spectroscopy measurements may be better suited for higher throughput applications during active monitoring of biomanufacturing processes.

[0087] In addition, infrared spectroscopy measurements are generally less sensitive to changes in environmental conditions that occur when the measurement is made. For example, relative to Raman measurements, infrared spectroscopy measurements are generally less sensitive to changes in temperature, humidity, and ambient light. Therefore, chemometric models based on such measurements, once calibrated, can be used under a wider variety of conditions and can be more easily translated from the laboratory to commercial, high-throughput manufacturing operations.

[0088] Infrared spectroscopy and measurement systems

[0089] Figure 1 1 is a schematic diagram showing an embodiment of a measurement system 100 for measuring infrared spectral information of a process solution in a biomanufacturing system. The system 100 includes a flow cell 106 located between fluid conduits 102 and 104. Fluid 122 enters the flow cell 106 from the conduit 102 and flows along the Figure 1 The flow is directed through the pool 106 as indicated by the arrow and exits the pool 106 into the conduit 104 .

[0090] In some embodiments, such as Figure 1As shown, the flow cell 106 includes an attenuated total reflectance (ATR) interface for measuring infrared spectral information in an ATR configuration / mode. Figure 1 1, the ATR interface is implemented as a prism 108, which forms part of the inner cavity within the cell 106. A first optical fiber 110 is optically coupled to a first surface 124 of the prism 108 and to a light source 112. A second optical fiber 114 is optically coupled to a second surface 126 of the prism 108 and to a detector 116. The radiation source 112 and the detector 116 are electrically connected to a controller 120.

[0091] During operation, the controller 120 activates the radiation source 112 to generate incident radiation 130. The incident radiation 130 is coupled into the first optical fiber 110 and is transmitted by the first optical fiber 110 through the first surface 124 into the prism 108. After entering the prism 108, the incident radiation 130 propagates toward the surface 128 of the flow cell 106.

[0092] The angle of incidence θ of the incident radiation 130 relative to the normal to the surface 128 and the material forming the flow cell 106 are selected such that when the incident radiation 130 reaches the surface 128, the incident radiation undergoes total internal reflection at the surface 128, thereby generating reflected radiation 132. The reflected radiation 132 is coupled into the second optical fiber 114 through the surface 126, propagates through the second optical fiber 114, and is detected by the detector 116. The detector 116 converts the reflected radiation 132 into spectral information and sends the spectral information to the controller 120. As will be explained in more detail below, the controller 120 analyzes the spectral information to determine information about one or more components of the fluid 122 and / or one or more process conditions associated with the biomanufacturing process that produced the fluid 122.

[0093] Radiation source 112 can include a variety of different radiation sources. In some embodiments, for example, radiation source 112 includes one or more light emitting diodes (LEDs). Typically, radiation source 112 generates incident radiation 130 in the infrared region of the electromagnetic spectrum. For example, incident radiation 130 typically includes radiation at one or more wavelengths between 750 nm and 500 microns.

[0094] In some embodiments, radiation source 112 generates incident radiation 130 at multiple different wavelengths. For example, radiation source 112 can generate incident radiation 130 in 3 or more (e.g., 5 or more, 7 or more, 10 or more, 15 or more, 20 or more, or even more) different, spectrally separated "bands" or "lines" (each having a center wavelength in the infrared region of the spectrum).

[0095] In some embodiments, radiation source 112 generates broadband incident radiation 130 over a range of wavelengths in the infrared spectral region. For example, incident radiation 130 can have a full width at half maximum bandwidth of 100 nm or greater (e.g., 200 nm or greater, 500 nm or greater, 1 micron or greater, 5 microns or greater, 10 microns or greater, 50 microns or greater, 100 microns or greater, 200 microns or greater, 300 microns or greater).

[0096] Detector 116 can generally be implemented in a variety of ways. For example, in some embodiments, detector 116 can be a Fourier transform infrared (FTIR) spectrometer that receives reflected radiation 132 and generates spectral information of the reflected radiation. One suitable FTIR spectrometer for use as detector 116 is the Bruker FTIR (available from Bruker Optics, Billerica, MA) with a mercury cadmium telluride (MCT) infrared sensor, but many other FTIR spectrometers can be used.

[0097] The detector 116 may also be implemented using standard optical elements that spatially disperse the frequency components of the reflected radiation 132, thereby mapping the radiation frequency to a spatial coordinate direction and then analyzing the dispersed frequency components as a function of position to measure the radiation intensity. Suitable optical elements for spatially dispersing the frequency components of the reflected radiation 132 include gratings, prisms, diffractive optical elements, and adaptive modulators, such as liquid crystal-based optical modulators. After the frequency components of the reflected radiation 132 have been spatially dispersed, the radiation intensity as a function of frequency may be measured using a variety of detectors suitable for detecting infrared radiation, including detectors based on one or more of mercury cadmium telluride, indium antimonide, arsenic antimonide, and lead selenide, and / or detectors having sensor features based on quantum wells and / or quantum dots.

[0098] like Figure 1 As shown, in some embodiments, incident radiation 130 and reflected radiation 132 can be transmitted to and from the ATR interface (e.g., prism 108) via optical fibers 110 and 114. However, in certain embodiments, both incident radiation 130 and reflected radiation can be transmitted to and from the ATR interface using a single optical fiber. Figure 2 Shown is a schematic cross-sectional view of an optical fiber 200 comprising a first radiation transmitting core 210 and a second radiation transmitting core 214 surrounded by an opaque cladding material 252. Integrated into one end of the optical fiber 200 is an ATR interface implemented as a prism 208. The prism 208 is positioned such that the surface 250 can be optically coupled to Figure 1 The flow cell 106 replaces the prism 108.

[0099] During operation, incident radiation 130 generated by radiation source 112 is coupled into optical fiber core 210 and propagates through core 210 to integrated prism 208. The incident radiation enters prism 208 and undergoes total internal reflection from surface 250 of prism 208, thereby generating reflected radiation 132. Reflected radiation 132 is coupled into core 214 and propagates back through optical fiber 200. At the end of optical fiber 200 opposite prism 208, reflected radiation 132 is coupled to detector 116 and analyzed as discussed above.

[0100] In general, the foregoing examples represent only a subset of optical fibers and probes that may be used to transmit incident radiation 130 and reflected radiation 132 in system 100. A wide variety of optical fibers and probes may be used, including some commercially available components. An example of a suitable commercially available probe is Diamond ATR probe (available from Bruker Optics).

[0101] Additionally, in some embodiments, the incident radiation 130 and / or the reflected radiation 132 can propagate through free space rather than being transmitted in an optical fiber. Free space propagation can allow additional optical elements (such as filters, apertures, beam splitters, mirrors, and lenses) to be inserted into the optical path of one or both of the incident radiation 130 and the reflected radiation 132, thereby allowing further control over the characteristics of the two beams.

[0102] like Figure 1 and 2 As shown, in some embodiments, the ATR interface can be implemented as a prism having geometric features (e.g., vertex angles) selected to ensure total internal reflection of incident radiation 130 when the incident radiation encounters the interface between the prism material and the fluid 122. More generally, the ATR interface can also be implemented in other ways. Figure 3 3 is a schematic diagram showing another example of an ATR interface implemented as a trapezoid 308 (or alternatively, as a truncated prism). A surface 328 of the trapezoid 308 forms a portion of the inner cavity of the flow cell 106. In addition, a reflective coating 362 is provided on a surface 360 ​​of the trapezoid 308, which is opposite to the surface 328.

[0103] During operation, incident radiation 330 (e.g., from an optical fiber) is coupled into the trapezoid 308 through the surface 324. The incident radiation 330 is incident at a first location 372 along the interface between the fluid 122 and the surface 328 of the trapezoid 308. The incident radiation undergoes total internal reflection at the surface 328, forming reflected radiation 332. The reflected radiation 332 is reflected by the coating 362 on the surface 360 ​​and is again incident at a location 374 along the interface between the fluid 122 and the surface 328 of the trapezoid 308. The reflected radiation 334 emerges from the location 374, is reflected by the coating 362, and is incident at a location 376 along the interface between the fluid 122 and the surface 328. The reflected radiation 336 emerges from the location 376 and is coupled out of the trapezoid 308 through the surface 326 (e.g., into the optical fiber).

[0104] The geometric features of the trapezoid 308 ensure that the incident radiation 330 interacts with the fluid 122 multiple times through the surface 328 of the trapezoid 308, which is relative to the Figure 1 The single interaction measurement in the example of FIG. 3 increases the signal-to-noise ratio of the spectral information carried by the reflected radiation 336. In general, the trapezoid 308 can be configured so that the number of interactions between the incident radiation 330 passing through the surface 328 and the fluid 122 is 2 or more times (e.g., 3 or more times, 5 or more times, 7 or more times, 10 or more times, or even more times).

[0105] Typically, the ATR interfaces (such as prism 108 and trapezoid 308) used in system 100 are formed of relatively high refractive index materials to ensure that total internal reflection occurs at the interface (e.g., surface 128 and / or 328) between prism 108 and / or trapezoid 308 and fluid 122. Suitable materials that can form ATR interfaces such as prism 108 and / or trapezoid 328 include, but are not limited to, diamond, germanium, various thallium halides, zinc selenide, and silicon.

[0106] exist Figure 1In the embodiment of the present invention, the flow cell 106 is implemented as a flow cell connected between the conduits 102 and 104. Fluid 122, which may include any one or more of the various process fluids at various positions in the biomanufacturing system, flows through the pool 106 when measuring and analyzing infrared spectral information. The flow cell 106 can be positioned at various positions in the biomanufacturing system to allow the measurement of spectral information for product quality assessment and process control. For example, in some embodiments, the flow cell 106 can be positioned at the outlet port of a bioreactor (e.g., a perfusion bioreactor, such as a tangential flow perfusion bioreactor). Alternatively or additionally, the flow cell 106 can be positioned along the fluid flow path between the components of the biomanufacturing system, including being positioned between the bioreactor and the purification unit, between two purification units, and / or after the purification unit. The biomanufacturing system can generally include a single flow cell 106 or a plurality of flow cells 106, each of which is positioned in the biomanufacturing system to measure spectral information at different positions.

[0107] In some embodiments, infrared spectroscopy measurements are taken when fluid 122 is not flowing in cell 106 . Figure 4 1 is a schematic diagram showing a flow cell 106 positioned between two fluid conduits 102 and 104. A first valve 402 is electrically connected to the controller 120 and is positioned between the conduit 102 and the flow cell 106. A second valve 404 is electrically connected to the controller 120 and is positioned between the flow cell 106 and the conduit 104. During operation, the controller 120 opens the valve 402 to allow a portion of the fluid 122 to enter the flow cell 106. When this portion of the fluid 122 has entered the flow cell 106, the controller 120 closes the valve 402. With the valve 404 remaining closed, this portion of the fluid 122 is temporarily trapped within the flow cell 106. As discussed above, infrared spectral information of the fluid 122 is measured by the controller 120, wherein the fluid 122 remains stationary within the flow cell 106; and then the controller 120 opens the valve 404 to allow a portion of the fluid 122 to flow out of the cell 106. At the same time or later, the controller 120 may also open the valve 402 to allow a new portion of the fluid 122 to enter the flow cell 106. In this way, infrared spectral information can be measured from a non-flowing portion of the fluid 122, which may be useful when the flow rate of the fluid 122 through the cell 106 would otherwise disrupt or interfere with obtaining accurate, reproducible infrared spectral information from the fluid. Such a situation may occur, for example, when the fluid flowing through the cell 106 is highly turbid, resulting in scattering of the incident radiation 130 and / or interaction with bubbles in the fluid 122 and other flow-related inhomogeneities.

[0108] Although Figure 1One example of a flow cell 106 that may be used with system 100 is shown, but more generally, flow cells having a variety of different geometries may be used. Figure 5 is a schematic diagram showing a flow cell 506 positioned between conduits 102 and 104. The ATR interface forms part of a lumen or channel wall within the flow cell 506. Figure 5 As shown, fluid 122 flows from conduit 102 through cell 506 and into conduit 104. While fluid 122 is flowing, or alternatively while fluid 122 is stationary within flow cell 506 (e.g., as discussed above, while a valve coupled to controller 120 regulates the flow of fluid 122 into and out of flow cell 506), infrared spectroscopy measurements are made by ATR of incident radiation 130 at interface 128.

[0109] Although Figure 1 The flow cell 106 in FIG. 5 defines a nonlinear two-dimensional fluid flow path from the conduit 102 through the flow cell 106 and into the conduit 104, but the flow cell 506 defines a linear one-dimensional fluid flow path due to the location of the fluid ports 502 and 504 at either end of the flow cell 506. For process fluids flowing at relatively high rates within a biomanufacturing system, the flow cell 506 may provide certain advantages over the flow cell 106. Specifically, the flow cell 506 may allow for higher aggregate fluid flow rates to be maintained, thereby ensuring that infrared spectroscopy measurements do not introduce a flow rate bottleneck into the continuous biomanufacturing process.

[0110] As discussed above, certain measurement systems disclosed herein are configured to perform infrared spectroscopy measurements using an ATR-based geometry. The ATR geometry takes advantage of the mismatch in refractive index between the material forming the ATR interface (eg, prisms 108 and 308 ) and the fluid 122 . Figure 6 1 is a schematic diagram showing a ray 602 of incident radiation 130 incident on a surface 128 of a prism 108. The surface 128 forms a boundary between the material forming the prism 108 (typically a material having a relatively high refractive index at the wavelength of the ray 602) and the fluid 122 (which typically has a relatively small refractive index). Snell's law describes the relationship between the angle of incidence and the angle of reflection, θ, at the interface. i and θ r The refractive index n of the prism and fluid p and n f The relationship between them is:

[0111] n p sinθ i =n f sinθ r [1]

[0112] With np The value relative to n f As the value of sinθ increases, r The value of increases to maintain the relationship in equation (1). In other words, the greater the refractive index mismatch between prism 108 and fluid 122, the greater the sinθ r The larger the value, the greater the refraction angle θ. r Increase. For n p With n f The incident angle θ can be i Choose so that sinθ r =1. That is, the refracted light 604 propagates in a direction tangent to the surface 128. This angle of incidence is called the critical angle θ c .

[0113] For angles of incidence greater than the critical angle, no refracted ray 604 is produced at the surface 128. Instead, the incident ray 602 undergoes total internal refraction at the surface 128, producing a reflected ray 606. However, at the surface 128, the evanescent field extends a short distance (commonly referred to as the penetration depth) beyond the interface and into the fluid 122. No energy flows across the interface. Nevertheless, the evanescent field interacts with the fluid 122 (and the components of the fluid) and, through this interaction, changes the spectral characteristics of the field. As a result, by analyzing the changes in the spectral characteristics of the reflected ray 606, information about various components (such as analytes and manufacturing byproducts) and fluid properties can be extracted.

[0114] In some embodiments, using an ATR geometry to measure infrared spectral information can provide certain advantages. For example, relative to transmission mode infrared spectroscopy measurements, the ATR geometry involves a penetration depth into the interrogation fluid through the evanescent field, which is relatively short compared to the conventional absorption path length used in transmission mode experiments. In general, the longer the path length, the greater the reduction in measurement sensitivity. Therefore, in some embodiments, infrared spectroscopy measurements based on ATR can be performed with higher sensitivity than transmission mode infrared absorption measurements.

[0115] Nonetheless, in some embodiments, chemometrics-based analysis may be performed on infrared spectral information obtained through transmission mode measurements. Figure 7 Shown is a schematic diagram of a measurement system 700 for use with a biomanufacturing system. The system 700 measures infrared spectral information of a fluid 122 using a transmission mode measurement geometry.

[0116] Fluid 122 flows from conduit 102 through flow cell 106 and into conduit 104. Controller 120 activates radiation source 112, generating incident radiation 130. Incident radiation 130 propagates through optical fiber 110 and couples through a window (the window forms part of the wall of flow cell 106), and enters a lumen or channel within flow cell 106. Once in the flow cell, at least a portion of incident radiation 130 is absorbed by one or more analytes or other components of fluid 122. The unabsorbed portion of incident radiation 130 emerges through a second window in flow cell 106 as transmitted radiation 702 and propagates through second optical fiber 114 to detector 116. An electrical signal generated by detector 116 encoding infrared spectral information carried by transmitted radiation 602 is transmitted to controller 120 for analysis.

[0117] Chemometrics-based analysis of infrared spectral information

[0118] Predictive, chemometric model-based spectroscopic analysis for process analysis of manufacturing small molecule drug substances has achieved some success. However, such methods have not been applied to the development of larger biotherapeutics (such as antibody-based drugs) because of the large size, complexity and heterogeneity of such products. In addition, the biomanufacturing processes for such substances are more complex than those for small molecule drugs and produce process solutions containing a wide range of analytes and other substances. To date, these complexities have advised against the application of chemometric methods to process control of biomanufacturing of large biomolecules (such as antibody-based drugs and nucleic acid-based products).

[0119] However, it has been found that despite the above complexities, chemometric model-based spectroscopic analysis can generate values ​​for various process-related parameters and product attributes associated with biomanufacturing of large biomolecules. These parameters and attributes can be used for real-time or near real-time process feedback and control. Specifically, vibrational spectral information from infrared reflection (or absorption) measurements is particularly suitable for analysis based on chemometric models because the spectral information provides a rich data set with prominent features such as stretching and bending resonances that are directly related to specific analytes and byproducts. More generally, spectral information encodes a complex set of responses of process fluid components to incident infrared radiation, and multivariate data analysis tools can decode these responses and, for certain parameters, can be used predictively with high accuracy.

[0120] In this section, methods and systems for constructing chemometric models and applying them to infrared spectral information are discussed. The methods and systems are particularly suitable for generating values ​​of various product attributes and parameters associated with process fluids generated during continuous biomanufacturing operations, and can be used to monitor such operations and perform process control by adjusting various biomanufacturing process inputs and characteristics. In particular, the methods can be used for real-time or near real-time biomanufacturing process feedback and control.

[0121] As will be described in more detail below, the multivariate data analysis methods discussed herein are used to separate the spectral information responses due to different analytes and quality attributes. After constructing a chemometric model for each analyte and / or quality attribute from data obtained from a reference method (such as chromatography or mass spectrometry), a single set of spectral information (e.g., infrared absorbance or reflectance spectrum) can be obtained, and the spectral information can be processed using the chemometric model to predict the quantitative values ​​of each quality attribute corresponding to the model from the same set of spectral information. Because the model operates on the same set of information, the predictive determination of quality attribute values ​​can occur in real time or near real time.

[0122] The chemometric model-based analytical methods discussed in this section operate on infrared spectral information to determine process-related parameter values. The infrared spectral information can be measured as described in the previous sections or by using other methods. In general, infrared spectral information corresponds to vibrational spectral information, and more specifically, corresponds to the vibrational spectrum of the process fluid (and the components contained in the fluid). The vibrational spectrum is a two-dimensional data set that includes intensity values ​​as a function of the wavelength or frequency of the radiation, where the range of wavelengths or frequencies falls within the infrared portion of the electromagnetic spectrum and therefore corresponds to typical wavelengths or frequencies of different vibrational modes of the components of the process fluid.

[0123] More generally, as used herein, vibrational spectroscopy information includes any data set representing the vibrational response (e.g., different vibrational modes) of components of a process fluid. The vibrational spectroscopy information may be encoded as a conventional vibrational spectrum, or in a different form representing similar information content as a vibrational spectrum.

[0124] Figure 8 800 is a flowchart showing a series of example steps that can be performed to analyze vibrational spectroscopic information (e.g., vibrational spectra) of a process fluid and its associated components to determine various property values ​​of the fluid and its associated components. The various steps shown in flowchart 800 can be performed in an automated manner by a system controller (e.g., controller 120) to perform the analysis.

[0125] In a first step 802, a calibrated and validated chemometric model is obtained for each property value to be determined. In some embodiments, the chemometric model (which typically consists of calibration coefficients describing the functional relationship between the property value of the process fluid and the spectral intensity values ​​at one or more wavelengths or frequencies) can be obtained by retrieving previously stored coefficient values ​​from a storage medium. Alternatively, in certain embodiments, the chemometric model is constructed and validated before being predictively used by the controller 120.

[0126] Fig. 9 902 is a flow chart showing a series of example steps that may be performed to construct and validate a chemometric model for a property of a process fluid based on vibrational spectroscopy information measured for the process fluid. In a first step 902, the controller 120 determines a set of characteristic model independent variables. In practice, the set of model variables corresponds to a set of independent vibrational frequencies within the vibrational spectroscopy information that predict the property for which the model is constructed. It should be understood that although the following discussion refers to "frequencies" within the vibrational spectroscopy information, the discussion may also refer equivalently to "wavelengths" given the interrelationship between wavelengths and frequencies in the vibrational spectroscopy information. That is, the method of defining a chemometric model in terms of a set of frequencies is equivalent to the method of defining a chemometric model in terms of a set of wavelengths.

[0127] The set of independent vibration frequencies can be determined in various ways. For example, in some embodiments, the set of vibration frequencies can be determined by performing principal component analysis on multiple calibration data sets, each calibration data set corresponding to vibrational spectral information associated with a process fluid having different, known property values ​​of interest. The principal component analysis determines how many independent frequencies reliably predict the property value on each calibration data set and the values ​​of the independent frequencies. The set of principal components corresponds to the set of frequencies that form the independent model variables in step 902. Methods for performing principal component analysis are described, for example, in Bro et al., "Principal component analysis," Anal. Methods 6:2812-2831 (2014) and Chatfield et al., "Principal component analysis," Introduction to Multivariate Analysis , Springer, pp. 51-87 (1980), the entire contents of each of which are incorporated herein by reference.

[0128] Next, in step 904, a chemometric model for the property of interest is constructed based on the set of independent variables from step 902. In general, a chemometric model can take a variety of functional forms. One such form is a linear model, in which the value of the property A of interest is expressed as a linear function of the set of independent variables:

[0129] A=a1Iυ1 +a2I υ2 +…+a i I υi +…+a n I υn [2]

[0130] where υ1…υ n is a set of n frequencies corresponding to the set of independent variables determined in step 902 (i.e., a set of n principal components), and I υ1 …I υn is a set of n intensity values ​​from the vibrational spectral information corresponding to each of the n frequencies. Parameters a1…a n is a collection of coefficients that effectively weights the contribution of individual intensity values ​​from the vibrational spectroscopy information to the predicted property A value.

[0131] Typically, as shown in equation (2), the chemometric model for the value of attribute A is a multivariate model, in which the value of attribute A depends on multiple independent variable values. The number of independent variables in the chemometric model can generally be selected as needed, for example, based on the results of a principal component analysis. Thus, for example, the value of attribute A in the chemometric model can be expressed as a function of one or more (e.g., two or more, three or more, four or more, five or more, six or more, seven or more, eight or more, ten or more, 12 or more, 15 or more, 20 or more, or even more) independent variables.

[0132] In some embodiments, the chemometric model has a more complex form than the linear form of equation (2). For example, certain chemometric models may be nonlinear, where the value of the attribute of interest A is expressed as a nonlinear function of a set of independent variables. Such a functional form of a chemometric model may be expressed as follows:

[0133] A=f1(I υ1 )+f2(I υ2 )+…+f i (I υi )+…+f n (I υn ) [3]

[0134] The function forms f j (I υj ) in I υj is linear or nonlinear. For equation (3) representing the nonlinear chemometric model of A, f j (I υj ) is at least one of I υj Non-linear function.

[0135] Usually, each f j (I υj ) Functional form in I υj can be linear or nonlinear. j (I υj ) is I υj In the case of a nonlinear function, f j (I υj ) can have any of a variety of different forms, such as, but not limited to: exponential form, logarithmic form, polynomial form, power law form, trigonometric form, hyperbolic form, and any combination of any of the foregoing functional forms. Such functional forms can generally be selected as needed during the construction of the chemometric model to ensure that the model's predictive ability for the value of the attribute A is sufficiently accurate. The subsequent discussion focuses on the linear example of equation (2), but it should be understood that similar principles also apply to nonlinear chemometric models. That is, the nonlinear chemometric model is similar to the linear chemometric model, with coefficients determined from a calibration data set, and then used to predict the quality attribute value from the measured vibrational (i.e., infrared) spectral information.

[0136] Returning to the linear model of equation (2), in the next step 906, after the model is constructed, the coefficient values ​​are determined based on one or more calibration data sets. The following discussion assumes that the chemometric model of attribute A is a linear model defined according to equation (2) above. If the model is different, the method discussed below can be used with slight modifications to determine the coefficient values.

[0137] Each calibration data set corresponds to a set of vibrational spectral information (e.g., vibrational spectra) of a process fluid having different, known values ​​of the property A. With respect to equation (2), for each calibration data set, A and I υ1 …I υn The values ​​of are known. Therefore, in step 906, the coefficient values ​​a1…a are determined in a self-consistent manner on each calibration data set. n A single set of , so that equation (2) determines the known values ​​of attribute A for each calibration data set as accurately as possible.

[0138] Various methods can be used to determine the coefficient values ​​a1…a in all calibration data sets. n In some embodiments, for example, partial least squares regression analysis is performed simultaneously on all calibration data sets so that the sum of squared error terms for the values ​​of attribute A is minimized. Methods for performing partial least squares regression are described, for example, in Haenlein et al., “ABeginner's Guide to Partial Least Squares Analysis,” Understanding Statistics3(4):283-297 (2004) and Sellin, “Partial Least Squares Analysis,” Int. J. Educational Research 10(2):189-200 (1986), the entire contents of each of which are incorporated herein by reference.

[0139] Next, in step 908, the coefficient values ​​a1 ... a1 determined in step 906 may optionally be verified against one or more additional calibration data sets. n to verify that the chemometric model for property A predicts the property value within an acceptable error. This verification step involves using the chemometric model to predict one or more values ​​of property A of the process fluid based on one or more sets of calibration data for the process fluid (each set of calibration data is a set of vibrational spectroscopy information). Since the value of property A for each process fluid (for which the calibration data set is measured) is known, the accuracy of the predicted value of property A generated by the chemometric model can be easily evaluated.

[0140] In optional step 910, the coefficient values ​​a1 ... a n The collection of may be stored in a storage medium for later retrieval by the controller 120. The process shown in the flowchart 900 ends at step 912.

[0141] In some embodiments, one or more calibration data sets and the measured vibrational spectral information may be processed before being used to construct a chemometric model and / or generate predicted quality attribute values ​​from the measured information. In general, various processing steps may be performed. In some embodiments, for example, one or more calibration data sets and / or measured spectral information may be baseline corrected. In certain embodiments, one or more data sets and / or measured information may be subjected to processing steps such as mean normalization and / or derivatization. For example, such steps may be performed using commercial analysis software such as (available from MathWorks, Natick, MA).

[0142] In certain embodiments, the chemometric model that predicts the value of attribute A can also be specific to certain conditions under which the spectral information is measured. For example, the model can be tied to a specific process fluid (e.g., the fluid eluted from a specific chromatographic column) for which the information is measured. Thus, in some embodiments, the methods disclosed herein include generating more than one chemometric model to predict the value of attribute A.

[0143] As an example, a first chemometric model may be generated to predict the concentration of A in a process fluid eluting from a first chromatography column in a biomanufacturing system, and a second chemometric model may be generated to predict the concentration of A in a different process fluid eluting from a second chromatography column downstream of the first column. Since the compositions of the fluids eluting from the two columns are different, the chemometric models may be different, but each is specifically generated to accurately predict the value of property A at a certain measurement location in the system. In the case where multiple models for property A are generated, each model may optionally be stored.

[0144] Back to Figure 8 After obtaining a calibrated chemometric model for each property of interest, vibrational spectral information of the process fluid is obtained in step 804. As an example, infrared reflectance measurements using ATR geometry (as discussed above) can be used to obtain vibrational spectral information, which can correspond to a vibrational spectrum of the process fluid and its associated components.

[0145] Next, in step 806, the vibrational spectroscopy information is analyzed using the first chemometric model obtained in step 802 to determine a value of a first property of interest for the fluid. As discussed above in conjunction with flowchart 900, determining the value of each property A involves using the intensity values ​​I measured at each independent frequency variable from the vibrational spectroscopy information. υ1 …I υn The set of coefficients a1…a determined for the stoichiometric model n The value of A is calculated from a set of . Because the calculation is deterministic (eg, whether the stoichiometric model is linear as in equation (2) or more complex), it is performed very quickly by the controller 120 .

[0146] After determining the first property value, control passes to decision step 808. If additional property values ​​are to be determined for the process fluid, a chemometric model for the next property of interest is selected in step 812, and control returns to step 806, where the next property value of interest is determined. In the event that a total of m property values ​​are determined for the process fluid, m different chemometric models are applied to the vibrational spectroscopy information in step 806.

[0147] A significant advantage of the methods and systems disclosed herein is that, due to the richness of vibrational spectral information, each of the m property values ​​can be obtained from the same set of vibrational spectral information. That is, no new measurements are performed to determine each property value. Instead, the vibrational spectral information is measured only once, and multiple property values ​​are determined from the same set of vibrational spectral information, each of which can be determined based on a set of multiple independent variables. Because information measurement is typically the most time-consuming step in the process of determining a property value, the elapsed time period for determining multiple property values ​​can be significantly shorter than a comparable time period associated with a method involving multiple measurement steps.

[0148] In some embodiments, for example, the time period over which vibrational spectral information is obtained and one or more property values ​​are determined from the vibrational spectral information using a chemometric model is 30 seconds or less (e.g., 25 seconds or less, 20 seconds or less, 15 seconds or less, 10 seconds or less, 5 seconds or less, 2 seconds or less) from the time when incident radiation 130 is incident on the surface of the ATR interface in contact with fluid 122.

[0149] If all property values ​​of interest are determined in step 808, control passes to step 810. In step 810, the controller 120 adjusts one or more process parameters of the biomanufacturing system in response to the measured process fluid property values. In general, a variety of adjustments can be made based on the nature of the determined property values ​​and the impact of those property values ​​on the biomanufacturing process. Some examples of measuring properties and making corresponding adjustments to the biomanufacturing system are discussed in subsequent sections.

[0150] Next, in decision step 812, if the property value of the process fluid is to be continued to be determined (e.g., for continuous process monitoring and control), control returns to step 804, and new vibrational spectroscopy information of the process fluid is obtained after an appropriate time interval. Alternatively, if the property value is not to be continued to be determined, control passes to step 814, where the process of flowchart 800 ends.

[0151] In general, the process shown in flowchart 800 can be used to determine any number of property values ​​from the same set of vibrational spectroscopy information. For example, in some embodiments, different chemometric models can be used to determine 1 or more property values ​​(e.g., 2 or more property values, 3 or more property values, 4 or more property values, 5 or more property values, 6 or more property values, 8 or more property values, 10 or more property values, 12 or more property values, or even more property values) from the same vibrational spectroscopy information.

[0152] Typically, the set of independent variables associated with a given chemometric model is different from the set of independent variables associated with other chemometric models used to calculate different property values ​​from the same set of vibrational spectral information. For each chemometric model, the set of independent variables associated with the model may include two or more (e.g., three or more, four or more, five or more, six or more, eight or more, 10 or more, 12 or more, 15 or more, or even more) major vibrational components.

[0153] However, because the set of independent variables is usually determined independently for each stoichiometric model, the set of main vibration components associated with different stoichiometric models may have no common members. Alternatively, in some embodiments, the set of main vibration components associated with different stoichiometric models may have 1 or more (e.g., 2 or more, 3 or more, 4 or more, 5 or more) common members.

[0154] Chemometric models of process fluid and component properties

[0155] Typically, chemometric models can be constructed to determine a wide variety of different quality attributes associated with biomanufacturing processes. These quality attributes are generally divided into several categories, including, but not limited to: product quality attributes, which are related to the purity, integrity, yield, form, and other attributes of the products of the biomanufacturing process; product-related impurities, which are related to the properties of different synthetic byproducts present in the process fluids generated during the biomanufacturing process; and process-related impurities, which are related to byproducts and other undesirable substances caused by the process conditions within the system.

[0156] Chemometric models can usually be constructed to be applied to the biomanufacturing of a large number of different substances. Some of the quality attributes associated with different categories of substances are different, while some are similar. For example, in some embodiments, the methods disclosed herein can be applied to the biomanufacturing of protein therapeutic substances (such as antibodies, peptides, enzymes, and other substances with amino acid chains). For such substances, chemometric models can be constructed to predict quality attribute values, including: (a) product quality attributes, which include concentration, aggregates, charge variant distribution, purity, glycan spectrum, identity and integrity; (b) product-related impurities, such as protein fragments; and (c) process-related impurities, such as host cell proteins, residual host cell DNA, residual column ligands (e.g., protein A) and other impurities, such as buffer components, surfactants, process additives (such as insulin, poloxamer, detergents, polysorbate 80) and other compounds from bioreactors and chromatographic / separation columns.

[0157] In certain embodiments, the methods disclosed herein can be applied to the biomanufacturing of nucleic acid drugs, including: DNA-based substances, such as DNA, plasmids, oligonucleotides, aptamers, and DNA enzymes (e.g., DNase); and RNA-based substances, such as RNA aptamers, RNA decoys, microRNAs, and small interfering RNAs. For such substances, chemometric models can be constructed to predict quality attribute values, including: (a) product quality attributes, including concentration, identity, integrity, and aggregates; (b) product-related impurities, such as nucleic acid fragments and nucleic acid variants; and (c) process-related impurities, such as residual column ligands and other impurities, such as buffer components, surfactants, process additives (e.g., insulin, poloxamer, detergents, polysorbate 80), and other compounds from bioreactors and chromatography / separation columns.

[0158] In some embodiments, the methods disclosed herein can be applied to the biomanufacturing of gene therapy drugs. For such substances, chemometric models can be constructed to predict quality attribute values, including: (a) product quality attributes such as concentration, aggregation, identity and integrity; (b) product-related impurities such as empty capsids, fragments and vector impurities; and (c) process-related impurities such as host cell proteins, host cell DNA, residual helper viruses, residual helper virus proteins, residual helper virus DNA and other impurities such as buffer components, surfactants, process additives (such as insulin, poloxamer, detergents, polysorbate 80) and other compounds from bioreactors and chromatography / separation columns.

[0159] The values ​​of any of the aforementioned quality attributes can be predictively generated by an appropriate chemometric model. Thus, the attribute A in equations (2) and (3) can represent any of the aforementioned quality attributes.

[0160] Furthermore, it should be noted that a chemometric model can be constructed to predictively generate values ​​for any combination of the above-described quality attributes from a single set of measured spectral information (e.g., infrared vibrational spectra). In general, values ​​for any combination of two or more (e.g., three or more, four or more, five or more, six or more, seven or more, eight or more, ten or more, or even more) of the above-described quality attributes can be generated from the same spectral information by a suitably constructed chemometric model.

[0161] In the following discussion, specific examples of quality attributes related to biomanufacturing of protein-based therapeutic substances are provided. Spectral measurements are made on the process fluid 122 (and its components). Quality attribute values ​​determined from the measured spectral information can be used for various purposes, including product quality assessment, biomanufacturing process adjustment, and process termination. Aspects of using quality attribute values ​​to adjust the biomanufacturing process by the controller 120 will be discussed later.

[0162] (i) Antibody concentration value

[0163] In certain embodiments, such as when the desired product of a biomanufacturing process is an antibody-based pharmaceutical product, the antibody concentration value of a process fluid at a specific location within the biomanufacturing system can be related to the total yield of the desired product. Thus, in some embodiments, the methods disclosed herein include obtaining and applying a chemometric model to determine the antibody concentration value of a process fluid.

[0164] The set of independent vibrational frequencies used in the chemometric model for antibody concentration values ​​may include frequencies within specific frequency ranges. By including frequencies within these specific ranges, prediction errors associated with determining antibody concentration values ​​from vibrational spectroscopy information may be reduced.

[0165] For example, in some embodiments, the set of major vibrational frequencies or components used in the chemometric model for antibody concentration values ​​may include frequencies in the wavenumber range 1100 cm -1 To 1595cm -1 With wave number range 1600cm -1 To 1700cm -1 In some embodiments, the set of major vibrational frequencies or components includes at least one corresponding to a wavelength in the wave number range of 1100 cm -1 To 1595cm -1 components in the vibrational spectrum, and at least one corresponding to a -1 To 1700cm -1 The components of the vibrational spectral information.

[0166] (ii) Extent of protein aggregation

[0167] In certain embodiments, such as when the desired product of a biomanufacturing process is protein-based, the extent of protein aggregation in a process fluid can provide important information about protein interactions and product yields. Thus, in some embodiments, the methods disclosed herein include obtaining and applying a stoichiometric model to determine the extent of protein aggregation in a process fluid.

[0168] The set of independent vibrational frequencies used in the stoichiometric model for the extent of protein aggregation may include frequencies within a specific frequency range. For example, in some embodiments, each of the major vibrational frequencies or components used in the stoichiometric model for the extent of protein aggregation corresponds to frequencies within the wavenumber range 1393 cm -1 To 1554cm -1 、Range 1600cm -1 To 1635cm -1 And range 844cm -1 To 1180cm -1 Vibrational spectral information in at least one of the.

[0169] In certain embodiments, at least one of the principal vibrational frequencies or component sets used in the stoichiometric model for the extent of protein aggregation corresponds to a frequency range of 1393 cm -1 To 1554cm -1 The vibrational spectrum information in the spectrum contains at least one main vibrational frequency or component corresponding to a frequency range of 1600 cm -1 To 1635cm -1 The vibration spectrum information in the spectrum is that at least one main vibration frequency or component corresponds to a frequency range of 844cm -1 To 1180cm -1 Vibrational spectroscopy information.

[0170] (iii) Amount of host cell protein

[0171] In certain embodiments, the amount of host cell protein in a process fluid in a biomanufacturing system can be used by the controller 120 to adjust manufacturing process parameters to increase product yield and reduce the formation of byproducts. Thus, the methods disclosed herein can include obtaining and applying a stoichiometric model to determine the amount of host cell protein in a process fluid.

[0172] The set of independent vibrational frequencies used in the stoichiometric model for the amount of host cell protein may include frequencies within a particular frequency range. For example, in certain embodiments, each of the principal vibrational frequencies or components used in the stoichiometric model for the amount of host cell protein may correspond to frequencies within the wavenumber range 1500 cm -1 To 1600cm -1 、Wavelength range 1600cm -1 To 1680cm -1 、Wave number range 1414cm -1 To 1489cm -1 and wave number range 1174cm -1 To 1286cm -1 Vibrational spectral information in at least one of the.

[0173] In some embodiments, at least one of the principal vibrational frequencies or components used in the stoichiometric model for the amount of host cell protein corresponds to a wavelength in the wavenumber range 1500 cm -1 To 1600cm -1 The vibrational spectrum information in the sample contains at least one main vibrational frequency or component corresponding to a wavelength in the wavenumber range 1600 cm -1 To 1680cm -1 The vibration spectrum information in the sample contains at least one main vibration frequency or component corresponding to a frequency in the wave number range 1414cm -1 To 1489cm -1 The vibration spectrum information in the spectrum is that at least one main vibration frequency or component corresponds to a wave number range of 1174cm -1 To 1286cm -1 Vibrational spectroscopy information.

[0174] (iv) Charge variant distribution

[0175] In certain embodiments, the controller 120 can use the charge variant distribution of a process fluid in a biomanufacturing system to adjust manufacturing process parameters. Thus, the methods disclosed herein can include obtaining and applying a chemometric model to determine the charge variant distribution in a process fluid.

[0176] It has been determined that certain frequency ranges of the major vibrational frequencies or components of such stoichiometric models can produce stoichiometric models that more accurately predict the charge variant distribution values. In some embodiments, for example, each of the major vibrational frequencies or components used in the stoichiometric model for the charge variant distribution corresponds to frequencies in the wavenumber range 1118 cm -1 Up to 1500cm -1 、Wavelength range 1120cm -1 To 1470cm -1 and wave number range 1187cm -1 To 1839cm -1 Vibrational spectral information in at least one of the.

[0177] In certain embodiments, at least one of the principal vibrational frequencies or components used in the stoichiometric model for the charge variant distribution corresponds to a frequency in the wavenumber range 1118 cm -1 Up to 1500cm -1 The vibration spectrum information in the sample contains at least one main vibration frequency or component corresponding to a frequency in the wave number range 1120 cm -1 To 1470cm -1 The vibration spectrum information in the spectrum is that at least one main vibration frequency or component corresponds to a wave number range of 1187cm -1 To 1839cm -1Vibrational spectroscopy information.

[0178] Alternatively, in some embodiments, the main vibrational frequencies or components used in the stoichiometric model for the charge variant distribution include: a first set of at least three frequencies or components corresponding to the wave number range 1118 cm -1 Up to 1500cm -1 The vibration spectrum information in the second group of at least three frequencies or components, which correspond to the vibration spectrum information in the wave number range of 1120cm -1 To 1470cm -1 and a third group of at least three frequencies or components corresponding to the vibration spectrum information in the wave number range 1187cm -1 To 1839cm -1 As discussed previously, among these major vibrational frequencies or components, at least one or more may be common to at least two groups, or alternatively, there may be none in common to any two or more groups.

[0179] Bioreactor-based analyte stoichiometric modeling

[0180] Integrated continuous biomanufacturing systems typically perform continuous capture and post-capture processing (i.e., purification, polishing, and filtration) of pharmaceuticals and other reactor-derived products. In both continuously operated systems and more conventional batch-based bioreactor systems, methods for assessing bioreactor conditions and adjusting various process parameters are important to ensure high yields. In addition, for large-scale manufacturing operations, and under relatively sensitive internal bioreactor conditions, it is desirable that methods for monitoring process parameters be highly automated, robust, and provide information that can be used to automatically adjust bioreactor conditions.

[0181] It has been shown that the specific cell line and chemical medium used under production conditions significantly affect the volumetric and specific productivity of the bioreactor. Specifically, with appropriate selection of cell line and bioreactor medium, viable cell density (VCD) and volumetric productivity have been sustained for relatively long periods of time. Figures 18A-18D The viable cell density of the perfusion bioreactor over a 45-day period ( Fig.18A ), cell viability ( Fig.18B ), harvest titer ( Fig.18C ) and specific productivity ( Fig.18D These data show that under the test conditions, the reactor can be filled with 100x10 6 After the initial reactor conditions reached equilibrium, the viable cell density, cell viability, harvest titer, and specific productivity remained relatively stable.

[0182] Different VCD, pH conditions, and perfusion rates can adjust the conditions within the bioreactor. Specifically, changes in any one or more of these operating parameters can change the levels of various intermediates and culture medium components (such as lactate, ammonium ions, glucose, and glutamine). Due to the importance of these intermediates in regulating production rates and affecting cell viability, the values ​​of many such quantities can be measured and adjusted relatively frequently to ensure that the bioreactor operates within a range of conditions that ensure stability and relatively high productivity. Figures 19A-19D The results show the glucose concentration ( Fig.19A ), glutamine concentration ( Fig.19B ), lactate concentration( Fig.19C ) and ammonium ion concentration ( Fig.19D ) changes.

[0183] Because of the impact of these and other cell culture medium components and process intermediates on cell viability and productivity, measurement of the values ​​of these and other components provides important information about internal bioreactor processes and can further provide data for automatic adjustment of bioreactor conditions that stray too far from specific target conditions. Figures 20A-20D It shows the glucose concentration ( Fig. 20A ), harvest titer ( Fig. 20B ), lactate concentration( Fig. 20C ) and ammonium ion concentration ( Fig.20D )'s daily at-line measurement. Figures 20A-20D The data in the table are obtained from daily manual measurements where the bioreactor medium is sampled and Figures 20A-20D Each quantity in was analyzed separately.

[0184] Unfortunately, manually measuring the individual values ​​of the aforementioned quantities can be very time consuming, which limits the rate at which measurements can be made. The slower the rate at which measurements are made, the longer the delay before corrective adjustments to bioreactor conditions can occur. Thus, if cell viability differs too significantly from ideal conditions, product yield may be adversely affected. In addition, the frequency of manually collecting and processing samples to measure the values ​​of these bioreactor culture medium components results in relatively high use of consumables associated with the sampling methods employed. Over time, the attendant costs of consumables can be high.

[0185] The online infrared measurement techniques and chemometric methods discussed previously can be directly applied to the measurement of components such as glucose, glutamine, lactic acid and ammonium ions based on bioreactor culture medium. Compared with manual sampling and analysis methods, the speed at which infrared spectral information can be obtained and the non-invasive manner in which infrared spectral information can be obtained make such measurements very advantageous. In addition, chemometric methods with validation models allow the extraction of multiple values ​​of quantities from a single set of infrared spectral information. In this way, the concentrations of glucose, glutamine, lactic acid and ammonium ions (and other culture medium components) can each be obtained from a single measured infrared spectrum analyzed using an appropriate chemometric model, which significantly reduces the number of spectral measurements performed, and thereby increases the rate at which bioreactor conditions can be adjusted in response to the measured values.

[0186] Integration and Regulation of Biomanufacturing Systems

[0187] The measurement system disclosed herein can be integrated with a biomanufacturing system to provide feedback control for various components and steps in the synthesis and purification processes of various biological products. The measurement system is usually implemented online between components of the manufacturing system, so that flowing or stationary solutions can be analyzed in real time without sampling or transfer. The measurement system can also be used more conventionally for samples extracted from reaction vessels, holding tanks or chromatographic columns before performing infrared spectroscopy measurements.

[0188] An integrated and fully continuous process for manufacturing therapeutic protein drugs and other substances may include, for example, providing a liquid culture medium containing a substantially cell-free recombinant therapeutic protein, which is then fed into a first multi-column chromatography system (MCCS1). The next step involves capturing the recombinant therapeutic protein in the liquid culture medium using MCCS1, and then continuously feeding the MCCS1 eluate containing the recombinant therapeutic protein into a second multi-column chromatography system (MCCS2), and using MCCS2 to purify and polish the protein. The resulting eluate from MCCS2 is considered to be a therapeutic protein drug. The process is integrated and can be run continuously from the liquid culture medium to the eluate from MCCS2 (i.e., a therapeutic protein drug).

[0189] Biomanufacturing systems are typically used to perform the above-described processes. For example, such systems may include MCCS1 (which includes an inlet) and MCCS2 (which includes an outlet). In these systems, the first and second MCCSs are in fluid communication with each other. These systems are also configured so that a fluid can be conveyed into the inlet, through the first MCCS and the second MCCS, and out of the manufacturing system through the outlet.

[0190] Such a system can provide continuous, time-efficient production of therapeutic drugs from liquid culture medium. For example, the elapsed time between feeding a fluid (e.g., liquid culture medium) containing a therapeutic protein into the first MCCS and eluting the therapeutic protein drug (containing the therapeutic protein) from the outlet of the second MCCS can be, for example, between about 4 hours and about 48 hours.

[0191] Fig.10 is a schematic diagram showing an example of a biomanufacturing system. System 1 includes a first MCCS, a four-column periodic countercurrent chromatography system (PCCS) 2, wherein three of the four columns (columns 3, 4, and 5) of the four-column PCCS 2 perform a unit operation of capturing a recombinant therapeutic protein from a fluid (e.g., a liquid culture medium substantially free of mammalian cells) containing the recombinant therapeutic protein, and one of the columns (column 6) of the PCCS 2 performs a unit operation of inactivating viruses present in an eluate from columns 3, 4, and 5 of the PCCS 2 containing the recombinant therapeutic protein. Columns 3, 4, and 5 may contain resins that utilize a protein A binding capture mechanism. Column 6 is capable of maintaining a fluid at a pH of about 3.75 for about 1 hour. PCCS 1 also has an inlet 7. Inlet 7 may be, for example, an orifice for receiving a fluid into PCCS 1.

[0192] The system 1 further includes a second MCCS, which is a PCCS 8, which includes three chromatography columns 9, 10, and 11 and a chromatography membrane 12. The columns 9, 10, and 11 in the PCCS 8 may contain cation exchange resins. The chromatography membrane 12 in the PCCS 8 may contain cation exchange resins. The PCCS 8 further includes a fluid conduit 13 disposed between the columns 9, 10, and 11 in the PCCS 8 and the chromatography membrane 12 in the PCCS 8. The PCCS 8 further includes an online buffer adjustment reservoir 14 in fluid communication with the fluid conduit 13 and configured such that a buffer contained in the online buffer adjustment reservoir 14 is introduced into a fluid present in the fluid conduit 13. The PCCS 8 further includes an outlet 15. The outlet 15 may be, for example, an orifice that allows fluid to be discharged from the PCCS 8.

[0193] The system 1 may further include a fluid conduit 16 disposed between the PCCS 2 and the PCCS 8. The system 1 may also include an online buffer regulation reservoir 17 in fluid communication with the fluid conduit 16, which is configured such that a buffer contained in the online buffer regulation reservoir 17 can be introduced into a fluid present in the fluid conduit 16. The system 1 may also include a filter 18 disposed in the fluid conduit 16 to filter the fluid present in the fluid conduit 16. The system 1 may also include a cut-off groove 19 disposed in the fluid conduit 16 and configured to contain any fluid in the fluid conduit 16 that cannot be easily fed into the PCCS 8.

[0194] The system 1 may further include a pump system 20 in fluid communication with the inlet 7. The pump system 20 may include a pump 21 for pushing a fluid into the inlet 7. The system 1 may also include a fluid conduit 22 disposed between the pump 21 and the inlet 7. The system 1 may also include a filter 23 disposed in the fluid conduit 22 to filter a fluid (e.g., a liquid culture medium) present in the fluid conduit 22. The system 1 may also include a cut-off groove 24 disposed in the fluid conduit 22, which is configured such that the cut-off groove 24 is in fluid communication with the fluid conduit 22 and is capable of storing any fluid present in the fluid conduit 22 that cannot enter the inlet 7.

[0195] The system 1 may further include a bioreactor 25 and a fluid conduit 26 disposed between the bioreactor 25 and the pump 21. A filtration system 27 may be disposed in the fluid conduit 26 to filter the liquid culture medium present in the fluid conduit 26 (eg, to remove cells therefrom).

[0196] The first MCCS (PCCS 2) includes an inlet through which a fluid (e.g., a substantially cell-free liquid culture medium) can be delivered into the first MCCS. The inlet can be any structure known in the art for such a purpose. It can include, for example, threads, ribs, or seals that allow for insertion of a fluid conduit, such that after the fluid conduit is inserted into the inlet, the fluid will enter the first MCCS through the inlet without a significant amount of fluid leaking out of the inlet.

[0197] The first MCCS includes at least two chromatographic columns, at least two chromatographic membranes, or at least one chromatographic column and at least one chromatographic membrane, and an inlet. For example, the first MCCS may include a total of four chromatographic columns, or three chromatographic columns and one chromatographic membrane, or any other exemplary MCCS described herein, or having one or more of any exemplary features (in any combination) of the MCCS described herein.

[0198] One or more chromatographic columns and / or one or more chromatographic membranes present in the first MCCS may contain one or more of various resins. For example, the resin contained in one or more of the one or more chromatographic columns and / or one or more chromatographic membranes present in the first MCCS may be a resin that utilizes a capture mechanism (e.g., a protein A binding capture mechanism, a protein G binding capture mechanism, an antibody or antibody fragment binding capture mechanism, a substrate binding capture mechanism, a cofactor binding capture mechanism, an aptamer binding capture mechanism, and / or a tag binding capture mechanism). The resin contained in one or more of the one or more chromatographic columns and / or one or more chromatographic membranes of the first MCCS may be a cation exchange resin, an anion exchange resin, a molecular sieve resin, or a hydrophobic interaction resin, or any combination thereof. Other examples of resins that can be used to purify recombinant therapeutic proteins are known in the art and may be contained in one or more of the one or more chromatographic columns and / or one or more chromatographic membranes present in the first MCCS. One or more chromatographic columns and / or chromatographic membranes present in the first MCCS may contain the same and / or different resins (e.g., any resin described herein or known in the art for purification of recombinant proteins).

[0199] The two or more chromatography columns and / or chromatography resins present in the first MCCS can perform one or more unit operations (e.g., capturing recombinant therapeutic proteins, purifying recombinant therapeutic proteins, polishing recombinant therapeutic proteins, inactivating viruses, adjusting the ion concentration and / or pH of a fluid containing a recombinant therapeutic protein, or filtering a fluid containing a recombinant therapeutic protein). In a non-limiting example, the first MCCS can perform unit operations of capturing recombinant therapeutic proteins from a fluid (e.g., a liquid culture medium) and inactivating viruses present in a fluid containing a recombinant therapeutic protein. The first MCCS can perform any combination of two or more unit operations described herein or known in the art.

[0200] One or more chromatographic columns and / or one or more chromatographic membranes present in the first MCCS can be connected or moved relative to each other by a switching mechanism (e.g., a column switching mechanism). The first MCCS may also include one or more (e.g., two, three, four, or five) pumps (e.g., automated, such as automatic peristaltic pumps). A column switching event can be triggered by detecting the level of a recombinant therapeutic protein in a fluid (e.g., input and / or eluate into and / or from one or more of the one or more chromatographic columns and / or chromatographic membranes in the first MCCS), a specific liquid (e.g., buffer) volume, or a specific elapsed time through the fluid passing through the first MCCS. Column switching generally refers to a mechanism that allows at least two different chromatographic columns and / or chromatographic membranes in a MCCS (e.g., two or more different chromatographic columns and / or chromatographic membranes present in a MCCS (e.g., a first or second MCCS)) to pass through different steps (e.g., equilibration, loading, elution, or washing) substantially simultaneously during at least a portion of the process.

[0201] The PCCS 2 as the first MCCS may include four chromatographic columns, wherein the first three columns perform a unit operation of capturing a recombinant therapeutic protein from a fluid (e.g., a liquid culture medium), and the fourth column of the PCCS performs a unit operation of inactivating viruses in a fluid containing the recombinant therapeutic protein. The PCCS as the first MCCS may use a column switching mechanism. The PCC system may utilize a modified chromatographic column capable of running up to, for example, four, five, six, seven, or eight columns, or more. system (GE Healthcare, Piscataway, NJ).

[0202] Column switching events can be triggered by detecting the concentration of a particular protein or other substance in a fluid eluting from a column of PCCS 2 or PCCS 8, flowing through a filter in an MCCS, contained in a cut-off channel in an MCCS, or flowing through a conduit in an MCCS (e.g., between MCCS 1 and MCCS 2). The measurement system disclosed herein can be used to measure the concentration of such proteins and communicate the concentration information to a controller in system 1, which initiates events such as column switching, filtration, and fluid transfer in system 1.

[0203] The first MCCS may be equipped with: one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) measurement systems configured to obtain infrared spectral information of the process fluid (e.g., system 100), one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) valves, one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) pH meters, and / or one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) conductivity meters. The first MCCS may also be equipped with a controller that executes an operating system that utilizes software (e.g., software based on Unicorn (GE Healthcare, Piscataway, NJ) or other software that implements similar functionality) to determine when column switching should occur (e.g., based on concentration information from infrared spectroscopy measurements, liquid volume, or elapsed time) and to affect (trigger) column switching events. The measurement system may optionally be placed at the inlet of one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) of the one or more chromatographic columns and / or the one or more chromatographic membranes in the first MCCS, and / or at the outlet of one or more chromatographic columns and / or the one or more chromatographic membranes in the first MCCS.

[0204] The first MCCS may further include one or more (e.g., two, three, four, five, six, seven, eight, nine, ten, eleven, twelve, thirteen, fourteen, fifteen, sixteen, seventeen, eighteen, nineteen, twenty, twenty-one, twenty-two, twenty-three, or twenty-four) online buffer conditioning reservoirs and / or buffer reservoirs. In other examples, the first MCCS may include one or more (e.g., two, three, four, five, or six) cut-off slots that can accommodate fluids that cannot easily enter one or more chromatographic columns and / or chromatographic membranes in the first MCCS. The systems described herein may contain one or more cut-off slots (e.g., the cut-off slots described herein) in the first and / or second MCCS. Other examples of the systems described herein do not include a cut-off slot in the first MCCS or the second MCCS, or do not include a cut-off slot in the entire system. Other examples of the systems include a maximum of one, two, three, four, or five cut-off slots in the entire system.

[0205] In some embodiments, the first MCCS may include a virus inactivation device. For example, referring to Fig.10In some embodiments, the first MCCS includes a virus inactivation device 6 (i.e., in place of the column 6 described above). The virus inactivation device 6 is configured to inactivate viruses and viral vectors used in biomanufacturing processes. In some embodiments, for example, the virus inactivation device 6 includes a mixing vessel. Alternatively, in some embodiments, for example, the device 6 includes a piston flow inactivation system. Various examples of these virus inactivation devices help eliminate active viruses and viral vectors from process fluids in the first MCCS.

[0206] The second MCCS includes at least two chromatographic columns, at least two chromatographic membranes, or at least one chromatographic column and at least one chromatographic membrane, and an outlet. For example, the second MCCS may include a total of four chromatographic columns, three chromatographic columns and one chromatographic membrane, or any other exemplary MCCS described herein, or may have one or more of any exemplary features (in any combination) of the MCCS described herein. The one or more chromatographic columns and / or one or more chromatographic membranes present in the second MCCS may have one or more of the following: any shape, size, volume (bed volume) and / or unit operation described herein. The resin contained in one or more of the one or more chromatographic columns and / or one or more chromatographic membranes present in the second MCCS may be a resin that utilizes a capture mechanism (e.g., a protein A binding capture mechanism, a protein G binding capture mechanism, an antibody or antibody fragment binding capture mechanism, a substrate binding capture mechanism, a cofactor binding capture mechanism, a tag binding capture mechanism, and / or an aptamer binding capture mechanism). Useful resins include, for example, cation exchange resins, anion exchange resins, molecular sieve resins, and hydrophobic interaction resins. The one or more chromatography columns and / or chromatography membranes present in the second MCCS may contain the same and / or different resins (eg, any resin described herein or known in the art for recombinant protein purification).

[0207] One or more chromatographic columns and / or one or more chromatographic membranes present in the second MCCS may perform one or more unit operations (e.g., any unit operation described herein or any combination of the unit operations described herein). In a non-limiting example, the second MCCS may perform a unit operation of purifying a recombinant therapeutic protein from a fluid and finely treating a recombinant therapeutic protein present in a fluid containing the recombinant therapeutic protein. In other non-limiting examples, the second MCCS may perform unit operations of purifying a recombinant therapeutic protein present in a fluid, finely treating a recombinant therapeutic protein present in a fluid, and filtering a fluid containing the recombinant therapeutic protein. In another example, the second MCCS may perform unit operations of purifying a recombinant therapeutic protein present in a fluid, finely treating a recombinant therapeutic protein present in a fluid, filtering a fluid containing the recombinant therapeutic protein, and adjusting the ion concentration and / or pH of a fluid containing the recombinant therapeutic protein. The second MCCS may perform any combination of two or more unit operations described herein or known in the art.

[0208] The second MCCS may also include one or more (eg, two, three, four, or five) pumps (eg, automated, such as automated peristaltic pumps).

[0209] One or more chromatographic columns and / or one or more chromatographic membranes present in the second MCCS may be connected or moved relative to each other by a switching mechanism (e.g., a column switching mechanism). A column switching event may be triggered by detecting the level of a recombinant therapeutic protein or other substance (measured by infrared spectroscopy and analyzed using a chemometric model as discussed above to determine the level of a recombinant therapeutic protein in a fluid passing through the second MCCS (e.g., input and / or eluate into and / or from one or more of the one or more chromatographic columns and / or chromatographic membranes in the second MCCS), a specific volume of liquid (e.g., a buffer), or a specific elapsed time.

[0210] The PCCS 8 forming the second MCCS may contain three columns (performing a unit operation of purifying a recombinant therapeutic protein from a fluid) and a chromatographic membrane (performing a unit operation of polishing a recombinant therapeutic protein present in a fluid). For example, the three columns that perform the unit operation of purifying a recombinant therapeutic protein from a fluid may contain, for example, cation exchange resins, and the chromatographic membrane that performs the unit operation of polishing may contain a cation exchange resin. The PCCS as the second MCCS may use a column switching mechanism. For example, the PCCS may utilize a modified HPLC system that is capable of operating up to, for example, four, five, six, seven, or eight columns, or more. Systems (GE Healthcare, Piscataway, NJ)

[0211] Similar to the first MCCS, the second MCCS may also be equipped with: one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) infrared spectroscopy measurement systems, one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) valves, one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) pH meters, and / or one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) conductivity meters. The one or more measurement systems transmit the measured concentration information of the protein or other substance in the fluid to the controller, which uses the concentration information to determine whether to trigger a column switching event. The second MCCS may be equipped with an operating system executed by the controller that receives the concentration information, which uses software (e.g., software based on Unicorn, GE Healthcare, Piscataway, New Jersey) to determine when a column switching event should occur (e.g., based on infrared spectroscopy measurements, liquid volume, or elapsed time) and initiate a column switching event. In examples where the second MCCS includes one or more infrared spectroscopic measurement systems, the measurement systems may optionally be placed at the inlets of one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) of the one or more chromatographic columns and / or one or more chromatographic membranes in the second MCCS, and / or at the outlets of one or more chromatographic columns and / or one or more chromatographic membranes in the second MCCS.

[0212] The second MCCS may further include one or more (e.g., two, three, four, five, six, seven, eight, nine, ten, eleven, twelve, thirteen, fourteen, fifteen, sixteen, seventeen, eighteen, nineteen, twenty, twenty-one, twenty-two, twenty-three, or twenty-four) online buffer conditioning reservoirs and / or buffer reservoirs. In other examples, the second MCCS may include one or more (e.g., two, three, four, five, or six) cut-off slots (e.g., any cut-off slots described herein) that can accommodate fluids that cannot easily enter one or more chromatography columns and / or chromatography membranes in the second MCCS.

[0213] The second MCCS includes an outlet through which the therapeutic protein drug can leave the system. The outlet may include, for example, threads, ribs, or seals that allow insertion of a fluid conduit, or a vial designed to hold or store the therapeutic protein drug. The outlet may contain a surface that can be used to seal a sterile vial or other such storage container on the outlet to allow the recombinant protein drug to flow directly into the sterile vial or storage container.

[0214] One or more infrared spectrometry systems as disclosed herein may also be positioned to measure the concentration of the protein drug (or another substance) flowing out of the outlet. This information may be transmitted to the MCCS controller, which may determine the purity of the substance based on the information.

[0215] The systems described herein may also include a fluid conduit disposed between the first MCCS and the second MCCS. One or more infrared spectroscopic measurement systems may be disposed along the fluid conduit to determine information (e.g., concentration information) about the fluid contained in (e.g., flowing through) the conduit. This information may be transmitted to the MCCS controller, which may determine whether to initiate a column switching event based on the information as discussed above.

[0216] Any fluid conduit described herein can be, for example, a tube made of, for example, polyethylene, polycarbonate, or plastic. The fluid conduit disposed between the first MCCS and the second MCCS can further include one or more of any combination of the following: one or more online buffer regulation reservoirs, which are in fluid communication with the fluid conduit and are positioned so that the buffer stored in the one or more online buffer regulation reservoirs is added to the fluid present in the fluid conduit; a cut-off groove (e.g., any one or more cut-off grooves described herein), which is in fluid communication with the fluid conduit and is positioned so that it can accommodate any excess fluid present in the fluid conduit that cannot be easily fed to the second MCCS; and one or more filters, which are disposed in the fluid conduit so that they can filter the fluid present in the fluid conduit (e.g., remove bacteria). Any online buffer regulation reservoir may contain, for example, a volume of buffer between about 0.5L and 50L (e.g., at a temperature of or below 50°C, 37°C, 25°C, 15°C, or 10°C).

[0217] The systems described herein may optionally include a fluid conduit disposed between the final chromatography column or chromatography membrane and the outlet in the second MCCS. The systems described herein may further include one or more filters in fluid connection with the fluid conduit disposed between the final chromatography column or chromatography membrane and the outlet in the second MCCS, such that the filters can remove, for example, precipitated material, particulate matter, or bacteria from the fluid present in the fluid conduit disposed between the final chromatography column or chromatography membrane and the outlet in the second MCCS.

[0218] Some examples of the systems provided herein also include a bioreactor in fluid connection with the inlet of the first MCCS.Any exemplary bioreactor described herein or known in the art can be used in the systems of the present invention.

[0219] Some examples of the systems provided herein also include a pump system. The pump system may include one or more of the following: one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) pumps (e.g., any pump described herein or known in the art), one or more (e.g., two, three, four, or five) filters (e.g., any filter described herein or known in the art), one or more (e.g., two, three, four, five, six, seven, eight, nine, or ten) UV detectors, and one or more (e.g., two, three, four, or five) break tanks (e.g., any break tank described herein). Some examples of the systems provided herein further include a fluid conduit (e.g., any exemplary fluid conduit described herein or known in the art) disposed between the pump and the inlet of the first MCCS. In some examples, this particular fluid conduit may include one or more (e.g., two, three, or four) pumps (e.g., any pump described herein or known in the art) and / or one or more (e.g., two, three, or four) break tanks (e.g., any exemplary break tank described herein), wherein these pumps and / or break tanks are in fluid connection with the fluid present in the fluid conduit.

[0220] Some examples of the systems described herein further include another fluid conduit connected to the fluid conduit between the pump and the inlet, wherein one end of the another fluid conduit is fluidly connected to the bioreactor and the other end is fluidly connected to the fluid conduit between the pump and the inlet. This other fluid conduit may include a filter that is capable of removing cells from liquid culture medium taken from the bioreactor (e.g., ATF cell retention system).

[0221] The aforementioned biomanufacturing system allows the continuous production of therapeutic protein drugs. For example, the system provided herein allows the yield percentage of recombinant therapeutic protein (from starting material, such as starting liquid culture medium) to be greater than about 70%, greater than about 80%, greater than about 82%, greater than about 84%, greater than about 86%, greater than about 88%, greater than about 90%, greater than about 92%, greater than about 94%, greater than about 96% or greater than about 98%. The system described herein can also cause the yield percentage of recombinant therapeutic protein (from starting material, such as starting liquid culture medium) to be between about 80% to about 90%, between about 82% to about 90%, between about 84% to about 90%, between about 84% to about 88%, between about 84% to about 94%, between about 82% to about 92%, or between about 85% to about 95%.

[0222] The systems described herein can also result in the production of therapeutic protein drug products containing a recombinant therapeutic protein at a concentration greater than about 1.0 mg / mL, for example, greater than about 15 mg / mL, greater than about 20 mg / mL, greater than about 25 mg / mL, greater than about 30 mg / mL, greater than about 35 mg / mL, greater than about 40 mg / mL, greater than about 45 mg / mL, greater than about 50 mg / mL, greater than about 55 mg / mL, greater than about 60 mg / mL, greater than about 65 mg / mL, greater than about 70 mg / mL, greater than about 75 mg / mL, greater than about 80 mg / mL, greater than about 85 mg / mL, greater than about 90 mg / mL, greater than about 100 mg / mL, greater than about 125 mg / mL, or greater than about 150 mg / mL.

[0223] As discussed above, in some embodiments, the first and / or second MCCS can be a periodic countercurrent chromatography system (PCCS). PCCS can, for example, include two or more chromatographic columns (e.g., three columns or four columns), which are switched to allow continuous elution of recombinant therapeutic proteins from two or more chromatographic columns. PCCS can include two or more chromatographic columns, two or more chromatographic membranes, or at least one chromatographic column and at least one chromatographic membrane. Column operation is usually composed of loading, washing, elution and regeneration steps. In PCCS, multiple columns are used to discretely and continuously run the same steps in a cyclic manner. Since the columns are operated in series, the flow through one column and the washing liquid from the column are captured by another column. This unique feature of PCCS allows the loading of resin to be close to its static binding capacity rather than dynamic binding capacity, which is typical in batch mode chromatography processes.

[0224] An example of the three-column switching technique used in a PCCS containing three columns is shown in Fig.11 One cycle is defined as three complete column operations that produce an elution pool from each of the three columns used in the column switching technique. Once all steps in the cycle are completed, the cycle restarts. Due to the continuous cycle and elution, the fluid entering the PCCS is processed continuously, while the elution of the recombinant therapeutic protein from each column is discrete and periodic.

[0225] To advance from one step in the PCCS cycle to another, as Fig.11 The exemplary cycle shown in FIG. 1 employs a column switching strategy. Fig.11 The three columns of the exemplary PCCS system shown employ two automatic switching operations per column, the first of which is associated with initial product breakthrough and the second coincides with column saturation. Determining when a column switching operation should occur is based on information about the concentration of the recombinant therapeutic protein in the eluate from each chromatographic column of the PCCS.

[0226] As discussed above, the infrared spectroscopy measurement system disclosed herein can be used to determine the concentration of a recombinant therapeutic protein in the eluate from a PCCS column. The concentration information, which acts as a feedback control for the biomanufacturing system, is transmitted to the MCCS controller, which initiates column switching after determining that the switch is authorized.

[0227] For example, during column loading, the PCC control system can determine the baseline concentration of the therapeutic protein material eluted from the column (which is typically a zero concentration) using the infrared spectroscopy measurement system discussed above. During active elution, as the protein material breaks through, the measured protein concentration increases (e.g., above the baseline concentration). The system continues to monitor the increasing protein concentration, and when the concentration reaches a predetermined threshold, the flow from column 1 is directed to column 2 rather than to waste. Nominally, this occurs at time t1.

[0228] As the feed continues into column 1, column 1 is eventually almost saturated with protein product. At this point, the measured protein concentration in the eluate has reached another predetermined value, which occurs at time t2. At this point, the MCCS controller switches the inlet feed to column 2.

[0229] Regardless of the feed product concentration and capacity, the above-mentioned column switching strategy allows for uniform loading of the column. Similar switching of columns can be implemented based on the detected recombinant protein levels in the eluate from each column. Column switching can also be based on the elapsed time or amount of fluid (e.g., buffer) passing through one or more chromatographic columns and / or chromatographic membranes in the first or second MCCS.

[0230] In addition to providing feedback information to control column switching events, the measurement systems disclosed herein can also provide feedback information for adjusting various other biomanufacturing steps and operating parameters. An example of such an adjustment is the controlled adjustment of buffer concentrations at various stages of a biomanufacturing process.

[0231] Typically, during the use of two or more MCCS in any process described herein, one or more (e.g., three, four, five, six, seven, eight, nine, ten, eleven, twelve, thirteen, fourteen, fifteen, sixteen, seventeen, eighteen, nineteen, twenty, twenty-one, twenty-two, twenty-three, or twenty-four) different types of buffers may be employed. As is known in the art, the one or more types of buffers used in two or more MCCS in the process described herein will depend on: the resins present in the one or more chromatographic columns and / or one or more chromatographic membranes of the two or more MCCS (e.g., the first MCCS and the second MCCS), the recombinant therapeutic protein, and the unit operation (e.g., any exemplary unit operation described herein) - which is performed by the specific one or more chromatographic columns and / or one or more chromatographic membranes of the two or more MCCS. During the use of two or more MCCS in any process described herein, the volume and type of buffer employed can also be determined by one skilled in the art (e.g., as discussed in more detail below). For example, the volume and one or more types of buffers employed during the use of two or more MCCS in any process described herein can be selected to optimize one or more of the following in the recombinant protein drug product: the total yield of the recombinant therapeutic protein, the activity of the recombinant therapeutic protein, the purity level of the recombinant therapeutic protein, and the removal of biological contaminants (e.g., the absence of active viruses, mycobacteria, yeast, bacteria, or mammalian cells) from a fluid containing the recombinant therapeutic protein.

[0232] The unit operation of adjusting the ion concentration and / or pH of a fluid containing a recombinant therapeutic protein can be performed using a MCCS (e.g., a first and / or second MCCS) that includes and utilizes a buffer adjustment reservoir (e.g., an in-line buffer adjustment reservoir) to add new or additional buffer solution to the fluid containing the recombinant therapeutic protein (e.g., between columns in a single MCCS, or after the last column in the second-to-last MCCS (e.g., a first MCCS) and before feeding the fluid containing the recombinant therapeutic protein to the first column of the next MCCS (e.g., a second MCCS). The in-line buffer adjustment reservoir can be of any size (e.g., greater than 100 mL) and can contain any buffer solution (e.g., a buffer solution having one or more of the following: an increased or decreased pH compared to the fluid containing the recombinant therapeutic protein; an increased or decreased ion (e.g., salt) concentration compared to the fluid containing the recombinant therapeutic protein; and / or an increased or decreased concentration of an agent that competes with the recombinant therapeutic protein for binding to a resin present in at least one chromatographic column or at least one chromatographic membrane of the MCCS (e.g., a first or second MCCS)).

[0233] In some embodiments, the MCCS controller determines that the amount of the buffer solution added to the process fluid is based on the concentration information of the process fluid components obtained from the infrared spectroscopy measurement as discussed above. For example, the solute used for such measurement can be a buffer solution component or a process fluid component, and its concentration is related to the fluid buffer composition, the pH of the process fluid, and / or the ionic strength of the process fluid. The measurement of the concentration information of the component is provided to the MCCS controller as feedback information, which uses the feedback information to determine when and how much one or more buffer solutions are discharged into the process fluid. The infrared spectroscopy measurement system can usually be positioned at any position in the biomanufacturing system for measuring the process fluid to provide the MCCS controller with feedback information related to the buffer.

[0234] In certain embodiments, the antibody concentration information of the process fluid can be used to control the rate at which the cell culture is introduced into the bioreactor. Specifically, by determining the antibody concentration value in the process fluid harvested from the bioreactor, the MCCS controller can adjust the rate at which the cell culture flows out of the bioreactor. Regulation in this manner allows control of the volumetric productivity of the cell density and specific productivity derived from the bioreactor. For a fixed perfusion rate, such regulation allows control of the antibody concentration in the process fluid so that the MCCS 1 will receive an approximately constant amount of product per unit time. In other words, regulating this property can be used to ensure that the product generation rate in the bioreactor remains approximately constant over a specific time period.

[0235] In some embodiments, determination of certain quality attributes associated with a process fluid may be used by the MCCS controller to determine whether the biomanufacturing system is operating within acceptable parameters, or whether during operation, the system is outside of one or more acceptable parameter ranges.

[0236] For each of the one or more quality attributes, a range of acceptable values ​​can be established through a calibration procedure. These ranges effectively establish operating conditions for the system under which the biological product is produced at an acceptable rate and purity level, while the production of byproducts and other undesirable substances is at an acceptably low level. When the system is operated outside of one or more ranges, product yield and / or purity may decrease, the rate / amount of undesirable substance production may increase, the rate of reagent consumption may increase, and / or other undesirable effects or conditions may occur.

[0237] The quality attributes determined for the process fluid at one or more locations within the system can be used to ensure that the system operates within an acceptable range of these operating parameters. If the determined values ​​of one or more quality attributes fall outside of the established acceptable range, the MCCS controller identifies that a potential fault condition exists.

[0238] To address a fault condition, the MCCS controller (or another system controller connected to the MCCS controller) can adjust any operating parameter of the biomanufacturing system to modify its operation, thereby also adjusting the quality attribute values ​​so that they fall within an acceptable range. Corrective actions of this nature ensure that the system can actively remain within an established set or range of operating conditions based on the feedback provided by the determined quality attribute values.

[0239] In certain embodiments, if the MCCS controller (or another system controller connected to the MCCS controller) determines that the system is so far from its acceptable range of operating conditions that returning the system to the acceptable range of conditions would be difficult or even impossible, or would result in other undesirable consequences, the controller may transmit a control signal to the bioreactor to suspend production and discharge its contents to waste. In such cases, the production process has deviated so far from the acceptable range of operating conditions for the system that effective corrective action is impractical or impossible. By simply draining the contents of the bioreactor, the system can save considerable time by restarting the production process, rather than attempting to adjust an ongoing production process that may have irreparably deviated from the acceptable range of conditions.

[0240] In addition, feedback can be provided to the MCCS controller (or another system controller) based on the measured values ​​of one or more bioreactor medium components (e.g., glucose concentration, glutamine concentration, lactate concentration, and ammonium ion concentration), which can then be used to adjust reactor conditions to ensure that cell viability, product yield, and other performance indicators are maintained within target ranges. The controller can adjust any one or more process parameters based on the values ​​of bioreactor medium components in a manner similar to the adjustments made based on the values ​​of product quality attributes and other measured quantities.

[0241] Example

[0242] The following examples are provided to further illustrate various aspects of the foregoing disclosure, but are not intended to otherwise limit any feature of the claims, or to limit any aspect of the embodiments, unless expressly stated otherwise.

[0243] To evaluate the methods and systems disclosed herein, FTIR spectroscopy using ATR geometry and partial least squares multivariate data analysis was used to develop a chemometric model to quickly and accurately determine multiple process physical and chemical attributes using a single set of (multi-attribute product quality (MAP-Q) vibrational spectral information. Protein A purified samples from antibody drug candidate X on multiple harvest days were analyzed for concentration, aggregates, charge variant distribution, and host cell protein (HCP) content using offline reference assays (chromatography and ELISA), and the same samples were subjected to FTIR ATR measurements. Data from offline reference assays and FTIR ATR infrared spectral information responses were used to construct a chemometric model using PLS multivariate data analysis, and the model was cross-validated to assess its accuracy.

[0244] Protein A purified, in-house produced harvest samples of antibody drug candidate X were used for this study. The monoclonal antibody used was of IgG4 subclass and expressed in a Chinese hamster ovary cell (CHO) mammalian expression system. Harvested samples (26 samples) were obtained from an alternating tangential flow perfusion bioreactor on different culture days. The harvest was purified using an AKTA Explorer system (available from GE Healthcare Life Sciences, Pittsburgh, Pennsylvania) using a 0.66 cm × 20 cm ID (6.8 mL) Protein A column filled with MabSelect SuRe LX resin.

[0245] Use equipped with Diamond ATR (Attenuated Total Reflection) Fiber Probe and MCT (Mercury Cadmium Telluride) Sensor from Bruker Near-line measurements of the samples were performed using an FTIR (Bruker Optics, Billerica, MA) spectrometer. Approximately 50 mL of each sample was placed on an ATR diamond crystal and measured using Bruker OPUS acquisition software (available from Bruker Optics, Billerica, MA) in the wavenumber range of 400 cm -1 Up to 4000cm -1 Spectra in (scan rate 10kHz, resolution 2cm -1 , and 32 scans per run). A reference spectrum was first recorded using a blank ATR cell built on the probe. Single beam spectra were obtained for all samples and separated from the background spectrum of air to display the spectra in absorbance units.

[0246] The infrared vibration information was preprocessed and chemometric partial least squares (PLS) models were constructed for each property of interest using MATLAB computing software (available from MathWorks, Natick, MA) and unscrambler camo software (CAMO Software, Inc., 33300 Egypt Lane, Magnolia, TX). During preprocessing, the 800-1800 cm -1 The FTIR spectra were offset by the average absorbance value between 1 and 2, and then baseline correction and area normalization were performed. Several preprocessing methods such as linear offset subtraction, line subtraction, vector normalization, min-max normalization, multivariate scatter correction, first-order derivative, and second-order derivative were evaluated to improve the root mean square error of cross validation (RMSECV) and the coefficient of determination (R) of the PLS model. 2 ). When building the PLS model, offline chromatographic / ELISA-based reference measurements of individual attributes were correlated with the pre-processed FTIR ATR spectral data.

[0247] To ensure validation of the calibration model, a cross-validation approach was applied to 20 protein A purification samples, where k-fold cross-validation was used to validate the individual FTIR ATR spectra of the samples, in which the data set was divided into k subsets and the k-fold model was trained and tested. Each time, one of the k subsets was used as a test set, and the other k-1 subsets were merged to form a training set. The average error of all k trials was calculated. The wavenumber (or frequency) range corresponding to each attribute value was optimized to achieve optimal multivariate statistics.

[0248] Off-line chromatography and ELISA were also used to measure the antibody concentration, aggregation, charge variant distribution and host cell protein (HCP) content of test sample to produce reference value. On Agilent 1100HPLC system (available from AgilentTechnologies company, Santa Clara City, California), 0.21 × 3cm ID (0.1mL) POROS protein A ID column (available from Applied Biosystems company, Foster City, California) was used to measure antibody concentration by protein A chromatography, and then UV measurement was carried out to eluate at 280nm. On Agilent 1100HPLC system, 0.78 × 30cm I.D. TSK glue G3000SWXL analytical SEC column (available from Tosoh Bioscience company, King of Prussia, Pennsylvania) equipped with 0.60 × 4cm ID TSK glue G3000SWXL guard column was used to analyze the percentage of the aggregation form of antibody in sample by size exclusion chromatography (SEC). 40mM sodium phosphate elution buffer in 150mM sodium chloride was used in isocratic mode, and UV absorbance at 280nm was then detected using a photodiode array detector. The HCP content of the samples was measured using a Chinese hamster ovary cell HCP ELISA kit from Cygnus technologies (available in Wrentham, Massachusetts) according to the manufacturer's manual. Samples were prepared at three dilutions, and then multiple measurements were performed on each sample.

[0249] Using iCE3 TM The iCE3 system (available from ProteinSimple, San Jose, CA) analyzes the charge variant distribution of antibodies in samples by capillary isoelectric focusing (cIEF). Each sample was prepared in a methylcellulose carrier ampholyte and allowed separation of charge variants (acidic and basic species) based on their pI under electrolytic conditions. TM The system's built-in full-column UV detector acquires the relative distribution of charge variants.

[0250] In infrared spectroscopy of protein molecules, the chemical structure of the molecule plays a major role, which determines the observed vibration frequencies through the strength of the vibrating bonds and the masses of the vibrating atoms. However, it can be difficult to unambiguously determine the chemical structure of a protein based purely on infrared spectroscopy due to the many overlapping bands that are typically present in the spectrum.

[0251] Nevertheless, it is often possible to detect changes in the chemical structure of a molecule based on changes in observed spectral bands. One such example is the detection of changes in the protonation state of protein side chains, which are often essential for protein function. The protonation state of many protein side chains is reflected in the infrared spectrum of the protein and can often be reliably inferred from the infrared spectral information.

[0252] The FTIR ATR spectrum of the protein A purified antibody sample is shown in Fig.12 , among which amide I (1600-1690cm -1 ) and amide II (1480-1575cm -1 The characteristic bands of amide I are clearly distinguishable. The amide I band is sensitive to the α-helix, β-sheet, turn and disordered conformations of the protein and its hydrogen bonding environment due to the C=O stretching vibration of the peptide backbone. Amide II originates from NH bending vibration and CN stretching vibration, and it is conformation sensitive.

[0253] In some cases, the amide II band corresponds to an out-of-phase combination of the NH in-plane bending mode and the CN stretching mode, with smaller contributions from the CO in-plane bending mode and the CC and NC stretching modes. In proteins, the amide II band is usually almost unaffected by side chain vibrations, but the correlation between protein secondary structure and frequency is not as simple as that of the amide I band, which helps to correlate the byproducts of proteins and provide valuable structural information and secondary structure predictions for the proteins.

[0254] 1300-1400cm -1 The amide III band in between is due to in-plane NH bending and CN stretching. The amide IV band is a very complex band resulting from a mixture of several coordinate shifts due to O=CN deformation, which mainly occur at 625-725 cm -1 The strength of hydrogen bonding in the secondary structure and the coupling between different switching dipoles of the peptide influences the absorption frequency in this region and can be used to quantify the protein from experimental spectral data.

[0255] In biological samples, each conformational entity contributes to the FTIR ATR spectrum of a molecule. The amide I band profile is a complex composite of several overlapping component bands representing different structural elements such as alpha helices, beta sheets, turns, and disordered or irregular structures.

[0256] In order to extract the compositional structural information encoded in those FTIR bands, the frequency range from 1100 to 1595 cm -1 and 1600 to 1700cm -1The FTIR spectral information in was used to calculate the PLS scale using the FTIR ATR spectra centered by the second derivative mean, excluding the absorbance caused by buffer components and water. The PLS model was constructed using the k-fold cross-validation method with 20 protein A purified samples of the antibody drug candidate X. The over-training of the model was also analyzed and the robustness of the model was evaluated.

[0257] Fig.13 is a graph showing the predicted antibody concentration values ​​calculated from the PLS model for antibody concentration and the corresponding measured antibody concentration values. The PLS model exhibits excellent correlation between the FTIR ATR predicted values ​​and the reference values ​​based on offline chromatography. The correlation coefficient R 2 The accuracy of the model was evaluated by predicting the antibody concentrations of 6 unknown samples using the developed calibration model. The results of unknown samples U-1 to U-6 are shown in Fig.17 in the table.

[0258] Aggregation of misfolded proteins can cause a shift in the key amide peaks in the FTIR ATR spectra. As a result, a band-narrowing Fourier self-deconvolution technique is used to estimate the FTIR range and position of discrete subcomponent absorption bands. Fourier self-deconvolution reduces the bandwidth, allowing the separation of overlapping component bands using a Gaussian function.

[0259] Due to the different hydrophobic environments, the vibration frequency of aggregated proteins drops to 1620-1625 cm -1 and often show a frequency shift of amide II (about 1540 cm -1 ). Therefore, 1393-1554cm -1 and 1600-1635cm -1 The FTIR region of 1180-844 cm was selected to develop the calibration model due to the COH vibration of proteins. -1 FTIR region to improve prediction accuracy. Fig.14A An example of a Fourier self-deconvoluted FTIR spectrum of a Protein A purification sample is shown.

[0260] Fig. 14B is a graph showing the PLS calibration model developed for aggregation (solid line) and measured aggregation values ​​(dots). The PLS calibration model shows excellent correlation between the FTIR predicted aggregation values ​​(%) with reference to the SEC values. RMSECV, R 2 and relative percentage difference (RDP) values ​​were 0.04, 0.97, and 5.8, respectively.

[0261] One of the challenges in constructing robust stoichiometric models for low concentration species such as HCPs is the effect of polymer electrolytes in the culture medium. -1 and 1005-110cm -1 The peak appearing at can be attributed to the polymer electrolyte. Fig.15A Shown is a set of mean-centered, baseline-corrected and area-averaged second-derivative FTIR ATR spectra used to build the PLS model of HCP.

[0262] To develop a PLS model for HCP quantification, a range of 1500–1600 cm -1 1600-1680cm -1 、1489-1414cm -1 and 1174-1286cm -1 All spectra are mean-centered, baseline-corrected, and area-averaged, as shown in Fig.15A shown.

[0263] Fig. 15B is a graph showing the model for HCP quantification (solid line) versus measured HCP values ​​(dots). A fairly good correlation was observed, with R 2 =0.89, RMSECV=71.1, and RPD=3.05.

[0264] To evaluate the PLS model for determining the charge variant distribution, the pre-processed FTIRATR spectra of the samples were used for k-fold cross validation using 20 samples for pre-main peak (acidic) species, main peak species, and post-main peak (basic) species to build three independent models, respectively. The over-training and robustness of the models were evaluated. The analysis of charge variants depends largely on the C-terminal lysine modification, which results in spectral patterns mainly between 1000 and 1850 cm -1 Therefore, a multiple scattering correction with the value from 1118 cm -1 Up to 1500cm -1 The PLS model of the main peak is constructed by analyzing the mean-centered spectrum of the second-order derivative analysis. Fig. 16B Middle (solid line). PLS statistics for the main peak model include an R of 0.99 2 , RMSECV of 0.00,1, and RPD of 12.4.

[0265] For analysis of pre-main peak substances, PLS levels up to 10 were used, starting at 1120 cm -1 To 1470cm -1 The FTIR region yielded the best PLS calibration model. The model is shown in Fig.16ARMSECV, R 2 and RPD values ​​are 0.00125, 0.9937, and 12.6, respectively.

[0266] For the analysis of the main peak, the peak from 1187 cm -1 To 1839cm -1 The PLS model is constructed in the FTIR region of Fig. 16C An excellent correlation with the measured values ​​was achieved from the reference cIEF method. RMSECV, R 2 The values ​​of and RPD are 0.00118, 99.59 and 15.7. The accuracy of the model was also evaluated by predicting the charge variant distribution of 6 unknown samples U-1 to U-6, and the results are shown in Fig.17 middle.

[0267] In order to evaluate the effectiveness of accurately determining the values ​​of bioreactor culture medium components and conditions using online infrared spectroscopy measurements and chemometric models, methods disclosed herein are used to measure glucose concentration, glutamine concentration, IgG concentration, lactic acid concentration, ammonia concentration, and molar osmotic pressure concentration. Specifically, online ATR infrared spectroscopy measurements are performed on the reactor culture medium, and the values ​​of each amount measured from the infrared spectroscopy measurements are predicted using a validated chemometric model for each of these amounts. In order to determine the accuracy of the infrared derived values, the reactor culture medium is also manually sampled and the values ​​of each amount are determined by separate analysis.

[0268] Figures 21A-21F is the glucose concentration of the bioreactor culture medium ( Fig.21A ), glutamine concentration ( Fig.21B ), IgG concentration ( Fig. 21C ), lactate concentration( Fig.21D ), ammonium ion concentration ( Fig.21E ), and osmolarity ( Fig.21F ) versus the predicted (or IR-derived) values ​​(solid line). For each measured quantity, good agreement between the true and predicted values ​​was achieved.

[0269] Figures 22A-22D It shows the glucose concentration ( Fig.22A ), harvest titer ( Fig. 22B ), lactate concentration( Fig. 22C ) and ammonium ion concentration ( Fig.22D ) versus predicted (dots). For each true value, multiple predicted values ​​were determined from multiple infrared spectra measurements, and Figures 22A-22DThe distribution of the points marked in the graphs shows the variability of the values ​​predicted based on the infrared spectral information. Although a relatively small number of outliers appear in the various graphs, most of the individually predicted values ​​are very consistent with the true values ​​of the various quantities, and the distribution of the values ​​determined by each group of infrared is consistent with the corresponding true values.

[0270] Figures 23A-23C is a graph showing the glucose concentration in the bioreactor culture at the later stages (on days 42 and 43) Fig.23A ), lactate concentration( Fig. 23B ) and ammonium ion concentration ( Fig.23C ) versus predicted (i.e., IR-measured, small dots) values. It is apparent from the figure that the IR-measured values ​​of these quantities remain consistently consistent with the true values, demonstrating that chemometric-based methods that rely on IR spectral information to determine relevant quantities of a variety of different media components remain highly accurate, even as cell cultures and media continue to produce product over relatively long periods of time.

Claims

1. A method comprising: obtaining a vibrational spectrum of a solution in a biomanufacturing system, wherein obtaining the vibrational spectrum comprises directing radiation through a first core of an optical fiber and incident on a first surface of an optical element distinct from the optical fiber and integrated into an end of the optical fiber such that at least a portion of the optical element is within a cladding of the optical fiber, and capturing in a second core of the optical fiber the incident radiation that undergoes total internal reflection from a second surface of the optical element; analyzing the vibrational spectrum using a first chemometric model to determine a first quality property value associated with the solution; analyzing the vibrational spectrum using a second chemometric model to determine a second quality property value associated with the solution; and At least one parameter of a purification unit of the biomanufacturing system is adjusted based on at least one of the first quality attribute value and the second quality attribute value.

2. The method according to claim 1, further comprising: The biomanufacturing system is used to produce at least one of a protein-based therapeutic substance, a nucleic acid-based drug, and a gene therapy drug.

3. The method of claim 2, wherein the protein-based therapeutic substance comprises at least one of a protein, a peptide, an antibody, and an enzyme.

4. The method of claim 2, wherein the nucleic acid-based pharmaceutical comprises at least one of DNA, plasmid, oligonucleotide, aptamer, DNA enzyme, RNA bait, micro RNA fragment and small interfering RNA fragment. The method according to claim 4 , wherein the aptamer is an RNA aptamer.

6. The method of claim 1, wherein for a biological product produced by the biomanufacturing system, the first quality attribute and the second quality attribute are each independently selected from a product quality attribute, a product-related impurity, and a process-related impurity.

7. The method of claim 1, wherein the second surface of the optical element contacts a flow cell containing a solution.

8. The method according to claim 7, further comprising: Attenuated totally reflected radiation is measured from the solution as the solution flows relative to the optical element.

9. The method of claim 1, wherein the optical element is a prism.

10. The method of claim 1, wherein the incident radiation comprises infrared radiation.

11. The method of claim 1 , wherein the first chemometric model includes a first set of principal vibrational components associated with the first mass attribute.

12. The method of claim 11, wherein the first stoichiometric model comprises at least three main vibrational components.

13. The method of claim 11, wherein the first stoichiometric model includes at least five major vibrational components.

14. The method of claim 11, wherein analyzing the vibrational spectrum using the first chemometric model comprises calculating the first mass attribute value based on the first set of principal vibrational components.

15. The method of claim 14, wherein calculating the first mass property value comprises determining the value as a linear function of the first set of primary vibration components.

16. The method of claim 11, wherein the second chemometric model includes a second set of principal vibrational components associated with the first mass attribute.

17. The method of claim 16, wherein the second stoichiometric model comprises at least three main vibrational components.

18. The method of claim 16, wherein the second stoichiometric model comprises at least five major vibrational components.

19. The method of claim 16, wherein the first set of principal vibrational components and the second set of principal vibrational components have no common members.

20. The method of claim 16, wherein analyzing the vibrational spectrum using the second chemometric model comprises calculating the second mass property value based on the second set of principal vibrational components.

21. The method of claim 20, wherein calculating the second mass property value comprises determining the value as a linear function of the second set of primary vibration components.

22. The method of claim 1, wherein the solution comprises a solution discharged from a purification unit of the biomanufacturing system.

23. The method of claim 22, further comprising: The solution in the purification unit is purified and then a vibrational spectrum of the solution is obtained.

24. The method of claim 22, wherein the purification unit comprises a chromatography column.

25. The method of claim 1, further comprising: The vibrational spectrum is obtained by measuring radiation from the solution as the solution flows between the first purification unit and the second purification unit of the biomanufacturing system.

26. The method of claim 25, wherein the first purification unit and the second purification unit each comprise a chromatography column.

27. The method of claim 1, further comprising: The vibrational spectrum is obtained by measuring radiation from the solution after it exits the final purification unit of the biomanufacturing system.

28. The method of claim 1, wherein the solution is a first solution, the method further comprising: obtaining a vibrational spectrum of a second solution in the biomanufacturing system; analyzing a vibrational spectrum of the second solution using the first chemometric model to determine a first quality property value of the second solution; and The vibrational spectrum of the second solution is analyzed using the second chemometric model to determine a second quality property value of the second solution.

29. The method of claim 28, wherein the first solution flows between a first purification unit and a second purification unit of the biomanufacturing system, and the second solution flows between the second purification unit and a third purification unit of the biomanufacturing system.

30. The method of claim 28, further comprising: The at least one parameter is adjusted based on at least one of the first quality property value and the second quality property value of the first solution and the first quality property value and the second quality property value of the second solution.

31. The method of claim 1, further comprising: The vibrational spectrum is obtained and the first quality attribute value and the second quality attribute value are determined within a period of 30 seconds or less starting from the time the radiation is incident on the solution.

32. The method of claim 31 , wherein the time period is 10 seconds or less.

33. The method of claim 31 , wherein the time period is 2 seconds or less.

34. The method of claim 1, further comprising: The steps of obtaining the vibrational spectrum, analyzing the vibrational spectrum using the first chemometric model, and analyzing the vibrational spectrum using the second chemometric model are repeated to determine a time series of first and second quality property values ​​associated with successive portions of the solution.

35. The method of claim 34, further comprising: The at least one parameter is adjusted based on at least one of the time series of the first quality attribute values ​​and the time series of the second quality attribute values.

36. The method of claim 16, further comprising: obtaining a collection of one or more calibration spectra representative of the solution; and The collection of calibration spectra is analyzed to determine the first set of principal vibrational components and the second set of principal vibrational components.

37. A method according to claim 36, wherein the first chemometric model includes a first set of coefficients associated with the first group of major vibrational components, and the second chemometric model includes a second set of coefficients associated with the second group of major vibrational components, and the method further includes determining the first set of coefficients and the second set of coefficients based on regression analysis.

38. The method of claim 16, further comprising: The vibrational spectrum is analyzed using a third chemometric model to determine a third quality property value associated with the solution.

39. The method of claim 38, wherein the third chemometric model includes a third set of principal vibrational components associated with the third mass attribute.

40. The method of claim 39, wherein the third stoichiometric model comprises at least three major vibrational components.

41. The method of claim 39, wherein the third stoichiometric model comprises at least five major vibrational components.

42. The method of claim 39, wherein analyzing the vibrational spectrum using the third chemometric model comprises calculating the third quality attribute value based on the third set of principal vibrational components.

43. The method of claim 42, wherein calculating the third mass property value comprises determining the value as a linear function of the third set of principal vibrational components.

44. The method of claim 42, further comprising: The at least one parameter is adjusted based on at least one of the first quality attribute value, the second quality attribute value, and the third quality attribute value.

45. The method of claim 38, further comprising: The vibrational spectrum is analyzed using a fourth chemometric model to determine a fourth quality property value associated with the solution.

46. ​​The method of claim 45, wherein the fourth chemometric model includes a fourth set of principal vibrational components associated with the fourth mass attribute.

47. The method of claim 46, wherein the fourth stoichiometric model comprises at least three major vibrational components.

48. The method of claim 46, wherein the fourth stoichiometric model comprises at least five major vibrational components.

49. The method of claim 46, wherein analyzing the vibrational spectrum using the fourth chemometric model comprises calculating the fourth quality attribute value based on the fourth set of principal vibrational components.

50. The method of claim 49, wherein calculating the fourth mass property value comprises determining the value as a linear function of the fourth set of principal vibration components.

51. The method of claim 45, further comprising: The at least one parameter is adjusted based on at least one of the first quality attribute value, the second quality attribute value, the third quality attribute value, and the fourth quality attribute value.

52. The method of claim 1, wherein at least one of the first quality attribute and the second quality attribute is selected from the group consisting of concentration, aggregates, charge variant distribution, purity, glycan profile, identity, and integrity.

53. The method of claim 1, wherein at least one of the first quality attribute and the second quality attribute is selected from the group consisting of protein fragments, nucleic acid fragments, nucleic acid variants, empty capsids, and vector impurities.

54. The method of claim 1, wherein at least one of the first quality attribute and the second quality attribute is selected from the group consisting of host cell proteins, residual host DNA, residual column ligands, impurity concentration, impurity amount, residual helper virus, residual helper virus proteins, and residual helper virus DNA.

55. A biomanufacturing system comprising: a bioreactor configured to produce a solution comprising a bioproduct; a purification unit configured to receive the solution; a radiation source configured to generate radiation; A detection device comprising: an optical fiber comprising a first optical fiber core and a second optical fiber core in a common cladding; an optical element distinct from the optical fiber and integrated into an end of the optical fiber such that at least a portion of the optical element is located within the common cladding, such that generated radiation propagates through the first optical fiber core and is incident on a first surface of the optical element, and the radiation undergoes total internal reflection from a second surface of the optical element and is coupled into the second optical fiber core; and a detector positioned to receive radiation from the second optical fiber core; and A system controller connected to the bioreactor and the detection device and configured to: receiving a measurement signal from the detection device, the measurement signal corresponding to information about the vibrational spectrum of the solution; analyzing the information using a first chemometric model to determine a first quality attribute value associated with the solution; analyzing the information using a second chemometric model to determine a second quality attribute value associated with the solution; and At least one parameter of the purification unit is adjusted based on at least one of the first quality attribute value and the second quality attribute value.

56. The system of claim 55, further comprising a flow cell positioned so that the solution passes through the flow cell, and the optical element is positioned so that the radiation is incident on the solution while the solution is in the flow cell.

57. The system of claim 55, wherein the biological product comprises at least one of a protein-based therapeutic substance, a nucleic acid-based drug, and a gene therapy drug.

58. The system of claim 57, wherein the protein-based therapeutic substance comprises at least one of a protein, a peptide, an antibody, and an enzyme.

59. The system of claim 57, wherein the nucleic acid-based pharmaceutical comprises at least one of DNA, plasmid, oligonucleotide, aptamer, DNA enzyme, RNA bait, micro RNA fragment, and small interfering RNA fragment.

60. The system of claim 59, wherein the aptamer is an RNA aptamer.

61. The system of claim 55, wherein for a biological product produced by the biomanufacturing system, the first quality attribute and the second quality attribute are each independently selected from a product quality attribute, a product-related impurity, and a process-related impurity.

62. The system of claim 56, wherein the detector comprises a total internal reflection sensor configured to measure attenuated totally reflected radiation from the solution.

63. The system of claim 62, wherein the total internal reflection sensor is integrated with a portion of the flow cell.

64. The system of claim 62, wherein the controller and the detection device are configured to measure attenuated total reflected radiation from the solution as the solution flows within the flow cell.

65. The system of claim 62, wherein the incident radiation comprises infrared radiation.

66. The system of claim 55, wherein the first chemometric model includes a first set of principal vibrational components associated with the first mass attribute.

67. A system according to claim 66, wherein the first stoichiometric model includes at least three major vibrational components.

68. The system of claim 66, wherein the first stoichiometric model comprises at least five major vibrational components.

69. The system of claim 66, wherein the controller is configured to analyze the vibrational spectrum using the first chemometric model by calculating the first mass property value based on the first set of principal vibrational components.

70. The system of claim 69, wherein the controller is configured to calculate the first mass property value as a linear function of the first set of primary vibration components.

71. The system of claim 66, wherein the second chemometric model includes a second set of principal vibrational components associated with the first mass attribute.

72. A system according to claim 71, wherein the second stoichiometric model includes at least three main vibrational components.

73. A system according to claim 71, wherein the second stoichiometric model includes at least five major vibrational components.

74. A system according to claim 71, wherein the first group of primary vibration components and the second group of primary vibration components have no common members.

75. The system of claim 71, wherein the controller is configured to analyze the vibrational spectrum using the second chemometric model by calculating the second mass property value based on the second set of primary vibrational components.

76. The system of claim 75, wherein the controller is configured to calculate the second mass property value as a linear function of the second set of primary vibration components.

77. The system of claim 55, wherein the solution comprises solution discharged from the purification unit.

78. The system of claim 77, wherein the controller is coupled to the purification unit and configured to purify a solution in the purification unit and then obtain information about a vibrational spectrum of the solution.

79. The system of claim 55, wherein the purification unit comprises a chromatography column.

80. The system of claim 55, wherein the purification unit is the first purification unit of the system, and the system further comprises: a second purification unit configured to receive the solution, Wherein, the detection device is positioned to measure radiation from the solution as the solution flows between the first purification unit and the second purification unit.

81. The system of claim 80, wherein the first purification unit and the second purification unit each comprise a chromatography column.

82. The system of claim 55, wherein the purification unit is the final purification unit of the system.

83. The system of claim 55, wherein the solution is a first solution, and the detection device is further configured to measure radiation from a second solution, and the controller is configured to: receiving a measurement signal from the detection device, the measurement signal corresponding to information about the vibrational spectrum of the second solution; analyzing information about the second solution using the first chemometric model to determine a first quality attribute value of the second solution; and The information about the second solution is analyzed using the second chemometric model to determine a second quality property value of the second solution.

84. The system of claim 83, wherein the purification unit is a first purification unit and the system comprises a second purification unit and a third purification unit, and wherein: The first solution flows between the first purification unit and the second purification unit; and The second solution flows between the second purification unit and the third purification unit.

85. The system of claim 83, wherein the controller is configured to adjust the at least one parameter based on at least one of the first quality attribute value and the second quality attribute value of the first solution and the first quality attribute value and the second quality attribute value of the second solution.

86. The system of claim 62, wherein the detection device and the controller are configured to measure radiation from the solution and determine the first quality attribute value and the second quality attribute value within a period of 30 seconds or less from the time the radiation is incident on the solution.

87. The system of claim 86, wherein the time period is 10 seconds or less.

88. The system of claim 86, wherein the time period is 2 seconds or less.

89. The system of claim 55, wherein the detection device and the controller are configured to repeat a plurality of steps of measuring radiation from the solution, analyzing information about a vibrational spectrum of the solution using the first chemometric model, and analyzing the vibrational spectrum using the second chemometric model to determine a time series of first and second quality attribute values ​​associated with successive portions of the solution.

90. The system of claim 55, wherein the controller is further configured to adjust the at least one parameter based on at least one of a time series of first quality attribute values ​​and a time series of second quality attribute values.

91. A system according to claim 71, wherein the controller is configured to obtain a set of one or more calibration spectra representative of the solution, and analyze the set of calibration spectra to determine the first set of principal vibrational components and the second set of principal vibrational components.

92. A system according to claim 91, wherein the first chemometric model includes a first set of coefficients associated with the first group of major vibrational components, and the second chemometric model includes a second set of coefficients associated with the second group of major vibrational components, and the controller is further configured to determine the first set of coefficients and the second set of coefficients by performing a regression analysis.

93. The system of claim 71, wherein the controller is configured to analyze information regarding the vibrational spectrum of the solution using a third chemometric model to determine a third quality property value associated with the solution.

94. The system of claim 93, wherein the third chemometric model includes a third set of principal vibrational components associated with the third mass attribute.

95. A system according to claim 94, wherein the third stoichiometric model includes at least three major vibrational components.

96. A system according to claim 94, wherein the third stoichiometric model includes at least five major vibrational components.

97. The system of claim 94, wherein the controller is configured to analyze information about the vibrational spectrum using the third chemometric model by calculating the third mass property value based on the third set of principal vibrational components.

98. The system of claim 97, wherein the controller is configured to calculate the third mass attribute value by determining the third mass attribute value as a linear function of the third set of primary vibration components.

99. The system of claim 97, wherein the controller is further configured to adjust the at least one parameter based on at least one of the first quality attribute value, the second quality attribute value, and the third quality attribute value.

100. The system of claim 94, wherein the controller is configured to analyze information regarding the vibrational spectrum using a fourth chemometric model to determine a fourth quality property value associated with the solution.

101. The system of claim 100, wherein the fourth chemometric model includes a fourth set of principal vibrational components associated with the fourth mass attribute.

102. The system of claim 101, wherein the fourth stoichiometric model comprises at least three major vibrational components.

103. The system of claim 101, wherein the fourth stoichiometric model comprises at least five major vibrational components.

104. The system of claim 101, wherein the controller is configured to analyze information about the vibrational spectrum using the fourth chemometric model by calculating the fourth quality property value based on the fourth set of principal vibrational components.

105. The system of claim 104, wherein the controller is configured to calculate the fourth mass property value by determining the fourth mass property value as a linear function of the fourth set of primary vibration components.

106. The system of claim 100, wherein the controller is configured to adjust the at least one parameter based on at least one of the first quality attribute value, the second quality attribute value, the third quality attribute value, and the fourth quality attribute value.

107. The system of claim 55, wherein at least one of the first quality attribute and the second quality attribute is selected from concentration, aggregates, charge variant distribution, purity, glycan profile, identity, and integrity.

108. The system of claim 55, wherein at least one of the first quality attribute and the second quality attribute is selected from the group consisting of protein fragments, nucleic acid fragments, nucleic acid variants, empty capsids, and vector impurities.

109. The system of claim 55, wherein at least one of the first quality attribute and the second quality attribute is selected from the group consisting of host cell proteins, residual host DNA, residual column ligands, impurity concentration, impurity amount, residual helper virus, residual helper virus proteins, and residual helper virus DNA.

Images