Detection of contamination in upstream and downstream bioprocesses by raman spectroscopy

Abstract

Methods according to the disclosure include detecting contamination in a biological fluid, comprising delivering a sample of a biological fluid from a bioreactor, biocontainer, or vessel during bioprocessing, taking Raman spectra data of the biological fluid, pre-processing the Raman spectra data; and analyzing the pre-processed data using clustering algorithms or qualitative mathematical tools or multivariate data analysis.

Description

DETECTION OF CONTAMINATION IN UPSTREAM AND DOWNSTREAM BIOPROCESSES BY RAMAN SPECTROSCOPYBACKGROUNDRelated Applications

[0001] The present application claims the benefit of priority of European Application No. 23306477.3, filed September 6, 2023, the entire contents of which is incorporated herein by reference.Field of the Disclosure

[0002] This disclosure relates to bioprocessing. More specifically, embodiments of the disclosure include the use of Raman spectroscopy and process analytics in-line, off-line, on-line, or at-line, to detect contamination of biological fluids, raw materials, and / or final products.Related Art

[0003] Bioprocessing is a generally complex process for producing biological products, e.g., organisms, cells, etc. Cell culture media (CCM) is used to feed organisms and cells during bioprocessing. Although there are many kinds of CCM, all can be affected by contaminants, e.g., unwanted viruses, bacteria, etc., at any stage. Bioprocessing is evolving and there is a need to understand and monitor what is happening in bioreactors during a cell culture run, during the media preparation step, incoming quality control step(s), or in a further step(s) of purification or formulation. Also, with the advent of longer continuous bioprocessing, such as perfusion processes, which is becoming more expensive, identifying and providing remedies, as soon as possible, during bioprocesses is becoming particularly more important. Bioprocesses may be stable processes with separate phases, typically for cell culture media preparation in upstream processes (USP) but also in downstream processes (DSP). Currently, sensors are used to monitor and characterize processing parameters such as pH, dissolved oxygen and carbon dioxide, temperature, cell count, turbidity, and conductivity, which are considered to be insufficient for monitoring bioprocesses. Raman probes have found use for monitoring biomolecules during these processes. At least one drawback during cell culturing, and in drug production, is contamination of the cell culture or of the media before starting a culture run or during purification steps, or of the final product. Such contamination has critical consequences, including, but not limited to, batch failure, i.e., does not pass Quality Control (QC)standards and regulations, as well as a waste of time and money. Moreover, severe safety issues exist for drug release / production and, consequently, for patients.

[0004] Currently, there are no direct, in-line and real-time methods for detecting contamination in a cell culture during a batch process, in liquid raw material (such as CCM) or in final product. In other words, before now, all methods were indirect or required incubation time.

[0005] With the embodiments described herein, it is now possible and advantageous to quantify cell culture nutrients, such as glucose and glutamine, among others and metabolites, e.g., ammonia, lactates, and others. It is also possible to monitor other cell properties, e.g., viable cell density (VCD) and total cell density, using the same probe. Furthermore, it is possible to quantify and directly correlate these data with the presence of bacteria and other living contamination within biological fluids, i.e., cell culture media and / or biological fluids during bioprocessing steps and / or final product. Some properties that portend contamination of biological fluids include: cell death due to a drop of dissolved oxygen, an increase in glucose consumption, unusual turbidity in the vessel, and others. Inline, at-line, off-line, and on-line methods for identifying contamination before and during bioprocessing of biological fluids, as well as products at final -fill represents an advance in the art.SUMMARY

[0006] Some embodiments of the disclosure, as set forth more completely in the claims, comprise off-line, on-line, real-time, at-line, and / or in-line sampling and analysis of biological fluids, raw materials, and / or final products, e.g., detection of contamination, in upstream and downstream bioprocesses by Raman spectroscopy without resorting to indirect methods. Some methods according to the disclosure comprise the use of a Raman probe in situ during bioprocessing to detect, in-line and in real-time, a possible contamination of the process, e.g., contamination by bacteria, fungi, and / or yeast, which represents a significant advance in the art. Data used for qualitative analysis or multivariate data analysis, as described herein, may be comprised of either or both of sterile and contaminated samples. Some embodiments of the disclosure comprise using Raman spectroscopy for detecting contamination, in biological fluids, raw materials, and / or final products during bioprocesses such as cell culture and / or continuous purification processes. Some embodiments of the disclosure comprise using Raman spectroscopy for detecting contamination during purification and / or during final filling steps. In some embodiments, classification / clustering algorithms are used to classify data and, in some embodiments, areused in conjunction with partial least squares and / or principal component analysis. Some embodiments of the subject matter comprise the use of an immersive probe. Some embodiments of the subject matter comprise at-line measurements using non-immersive, non-contact, and / or non-intrusive data acquisition, wherein the data can assess the presence or absence of bacteria.

[0007] This disclosure characterizes various properties of cell culture, particularly for mammalian cells cultured in bioreactors. Some embodiments of the disclosure comprise collecting Raman spectra, analyzing via multivariate data analysis (complex mathematical algorithms) and sending a signal or alert if contamination is detected. Various provisions, aspects, and novel and inventive features of the present disclosure, as well as details of exemplary embodiments thereof, will be more fully understood from the following description and drawings substantially as shown in or described with at least one of the figures and / or claims. So, the manner in which the features disclosed herein can be understood in detail, more particular descriptions of the embodiments of the disclosure, briefly summarized above, may be had by reference to the appended drawings. It is to be noted, however, that the appended drawings illustrate only typical embodiments of this disclosure and are therefore not to be considered limiting of its scope, for the described embodiments may admit to other equally effective surface treatments, methods, and / or materials. It is also to be understood that elements and features of one embodiment may be found in other embodiments without further recitation and that, where possible, identical reference numerals have been used to indicate comparable elements that are common to the figures. As used herein, the singular forms "a," "an," and "the" include plural referents unless the context clearly dictates otherwise. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which these embodiments pertain.BRIEF DESCRIPTION OF THE FIGURES

[0008] FIG. 1 depicts a flow chart for processing Raman spectra for detecting contamination within a biological fluid, according to embodiments of the disclosure;

[0009] FIG. 2 depicts a graph plotting data points of Raman spectra pre-processed by principal component analysis (PCA), demonstrating differences between contaminated cell culture media and non-contaminated cell culture media, according to some embodiments of the disclosure;

[0010] FIG. 3 depicts a graph showing pre-processed Raman spectra for glucose, lipids, and proteins, according to some embodiments of the disclosure;

[0011] FIG. 4 depicts Figures 4A and 4B which show comparisons between Raman spectra of contaminated and sterile cell cultures plotted on a PCA score plot, according to some embodiments of the disclosure;

[0012] FIG. 5 depicts a graph plotting bacterial concentration as a function of time, showing bacterial detection using a partial least squares (PLS) model, according to some embodiments of the disclosure;

[0013] FIG. 6 displays graphical representations of four modes of Raman measurement, according to some embodiments of the disclosure; and

[0014] FIG. 7 depicts a graph plotting a classification of data using a partial least squares- discriminant analysis (PLS-DA) clustering algorithm, according to embodiments of the disclosure.DETAILED DESCRIPTION

[0015] So the manner in which the features disclosed herein can be understood in detail, a more particular description of the embodiments of the disclosure, briefly summarized above, may be had by reference to the appended drawings. It is also to be understood that elements and features of one embodiment may be found in other embodiments without further recitation and that identical reference numerals are sometimes used to indicate comparable elements that are common to the figures.

[0016] The term “upstream” is defined as a process step that occurs before another step or process with respect to the direction of the workflow of a bioprocess. For example, upstream may indicate a process that occurs before a cell culturing process in a bioreactor, for example, cell culture media preparation. In some embodiments, the term “upstream” indicates a process for cultivating cells within a bioreactor. The term “downstream” is defined as a process step that occurs after another step or process with respect to the direction of the workflow of a chain of processes or unit operations, typically, a step occurring after cell culturing. From upstream to downstream, a series of unit operations in bioprocessing could include cell culture media preparation, cell culturing, capture chromatography, viral inactivation, an intermediate column chromatography step, ultrafiltration / diafiltration, a second chromatography step, a virus filtration step, and / or a purification / concentration step of a formulation.

[0017] The use of the term “partial least squares” indicates analyzing pre-processing data, comprising a statistical method based on a linear regression model by projecting observable variables X and predicted variables Y to new spaces, and finding the multidimensionaldirection in the X space that explains the maximum multidimensional variance direction in the Y space.

[0018] The use of the term “partial least squares-di scriminant analysis” indicates clustering algorithms, comprising a multivariate inverse least squares discrimination method used to classify samples, using predicted variables Y class labels as an input. Each sample data is projected on the components axis which maximizes the separation of the data in groups / clusters, each group / cluster corresponding to a class. The projection of a sample in either one of this group / cluster of the model defines its classification.

[0019] The use of term “principal component analysis” indicates qualitative mathematical tools, comprising a linear dimensionality reduction technique such that the directions (principal components) capturing the largest variation in the data can be easily identified, highlighting differences or similarities between samples by reducing complex spectral data to only two dimensions.

[0020] The use of the term “direct” indicates that a direct correlation between the presence of contamination and a signal measured. In this case, the signal measured can be the Raman spectra. The use of the term “indirect” indicates that a correlation can be made between the presence of contaminant and the consequences of this presence, meaning lower amounts or concentrations of properties than would be expected if no such contamination was present, using other qualities as a proxy for a measured signal. For example, in cell culture, such properties include a concentration of dissolved oxygen, glucose, glutamine, and other properties that are typically monitored during bioprocessing.

[0021] In some embodiments of processes, a drop in dissolved oxygen, despite sparging of the bioprocess, in 5% or less can indicate the presence of bacteria that is misappropriating the oxygen from the cells. In some embodiments of processes, a drop in dissolved oxygen, despite sparging of the bioprocess, in 10% or less can indicate the presence of bacteria that is misappropriating the oxygen from the cells. Similarly, drops in the concentration of glucose and / or glutamine can be correlated with the presence of a bacteria contaminant. In some embodiments of processes, a drop in the concentration of glucose can be observed, despite delivery of glucose into the bioprocess, can indicate the presence of bacteria that is misappropriating the glucose from the cells. For example, having to add more than 20 grams per liter per day in fed batch mode for e.g., in a 3L bioreactor, (although scaling is possible) would indicate that the bacteria is misappropriating the glucose. Raman spectroscopy can be used to compare the spectra obtained from samples taken during bioprocessing against spectra of reference samples,wherein a discrepancy indicates bacterial or other contamination. Deficiencies of these components, oxygen, glucose, glutamine, etc., contribute to cell death. Real-time monitoring and alerts can alleviate these problems, wasting of time, money, and resources.

[0022] In this disclosure, the term “sample” can refer to a small volume of biological fluid from a larger volume of this fluid, either taken at-line, off-line, but also in-line or on-line. In the case of in-line or on-line, the term sample indicate the volume of fluid where the measurement occurs when the probe is immersed in a larger volume.

[0023] The term “in-line” means that the sample measurements are taken in situ directly in the process.

[0024] The term “on-line” means that the sample measurements are made at a branch of the process, in a built-in branch (or "bypass" or in a recirculation loop or in an outlet tubing) into which samples of the substance to be measured are automatically fed.

[0025] The term “at-line” means that the sample measurements are not made directly in the process or at a branch of the process, but still in the laboratory close to the process in a measuring device external to the production system in the same time frame as the evolution of the process.

[0026] The term “off-line” means that the sample measurements are not made directly in the process or at a branch of the process, but temporally and / or spatially displaced from the process.

[0027] Turning now to the figures, FIG. 1 depicts a flow chart for processing Raman spectra for detecting contamination within a biological fluid, according to embodiments of the disclosure. At least one embodiment of the disclosure describes a method 100, as shown in FIG. 1, for detecting contamination in a fluid or testing sterility of a final product, raw material, or measuring bioburden, comprising delivering an on-line, in-line, off-line or at- line sample of a biological fluid from a bioreactor or bioprocess system or vial during bioprocessing at step 102; taking Raman spectra data of the biological fluid at step 104; pre-processing the Raman spectra data at step 106; and analyzing the pre-processed data using qualitative mathematical tools, clustering algorithms, or multivariate data analysis at step 108. Thereafter, the method 100 ends. The method 100 can be repeated automatically, continuously or periodically, and / or at irregular or regular intervals, e.g., once per hour, once per eight hours, once per day, etc. It should be noted that other steps can be added to the method 100. For example, the sample may be drawn from a bioprocess using a sampler, e.g., an automated sampler, and, further, from an aseptic modular automated sampler, which can be on-line, at-line, or in-line sampling technique and, further, any may be acontact measurement or a non-contact measurement. Furthermore, a microprocessor may be networked with the sampler, wherein a signal, indicating an out-of-specification condition, i.e., the presence of bacteria, fungi, and / or yeast within the bioreactor or sample measured, is sent to a computer, tablet, smartphone, etc., so that a human is alerted, for example, a visual and / or an audio alarm, optionally at step 110. Also, the signal may be adapted to take an automatic action within the bioreactor or bioprocess system, e.g., add a processing aid, such as a sanitizer, within the bioreactor through a port. The method 100 can be performed as incoming quality control for raw materials, during a bioprocessing run, such as a cell culturing process, or in processes downstream of the cell culturing step(s). It is to be noted that the method 100 can be performed whether sampled from a bioreactor bag, a delivery bag of raw materials, within a mixer bag, or other processes as are known to those in the art. It is to be further understood that the method 100 can be used during any unit operation within a bioprocessing run. For example, during or after clarification, filtering, virus inactivation, purification, etc., such as chromatography, anionic chromatography, cationic chromatography, mixed-mode chromatography, depth filtration, tangential flow filtration, final-fill processes, and / or the like.

[0028] A first experiment was conducted on 25 milliliter (mL) flasks containing either sterile or contaminated cell culture media. A spiking of differing concentrations of E. Coli BL21 bacteria was introduced within the media. The cell concentration of the bacteria ranged between 0.4 and 40,000 CFU / mL (colony forming units per milliliter). The initial concentration of bacterial solution was estimated using absorbance, while cascade dilutions were used to obtain the desired bacterial concentrations. Raman measurements were taken immediately after spiking using a Raman analyzer, such as the ProCellics™ Raman Analyzer and recording the data using software, such as on Bio4C® Raman PAT Software, both marketed by EMD Millipore Corporation, Burlington, MA, USA.

[0029] Once acquired, the Raman spectra was pre-processed in the software using specific algorithms, e.g., SNV normalization (standard normal variate), Savitzky-Golay derivative, and spectral selection. The spectral data can be corrected before or after from noises from the electronics of the sensor, from external straylight, and / or from cosmic rays. After, the pre-processed spectra can be analyzed using qualitative mathematical tools such as Principal Component Analysis (PCA). In this representation, each dot represents a pre- processed spectrum according to (at least) two dimensions representing the two (at least) main contributions.

[0030] FIG. 2 depicts a graph plotting data points of Raman spectra pre-processed by principal component analysis (PCA), demonstrating differences between contaminated cell culture media and non-contaminated cell culture media, according to embodiments of the disclosure. Raman spectroscopy has been effective in detecting contaminants in cell culture media, as shown in Figure 2. As demonstrated, the PCA score plot successfully highlighted differences between Raman spectra of contaminated CCM and sterile CCM with a limit of detection between 0.4 and 4 colony forming units (CFU) / mL. As previously stated, in-line, on-line, and / or real-time detection of contaminants in cell culture media using Raman spectroscopy has never been reported before. Hence, Raman spectroscopy demonstrates its direct ability to detect contaminations in cell culture media. In this representation, i.e., a map of weight, the correlation x / y does not matter. Rather, this graph helps to perform a qualitative analysis and identify clusters. For example, in Figure 2, there is one cluster of contaminated samples and one cluster of non-contaminated (sterile) samples. When analyzing an unknown sample, we measure its Raman spectra and plot it among the other Raman spectra on the PCA and we observe in which group it belongs to identify if this sample is contaminated or sterile.

[0031] In addition, Raman spectroscopy demonstrated its indirect ability to detect contaminations in cell culture. Cell cultures with four-bioreactor system were monitored in a second experiment by ProCellics™ Raman Analyzer in Multi-Channel Unit configuration (four probes in parallel) controlled by Bio4C® PAT Raman Software. Eight batches were generated in bioreactors, including three free contaminations, three controlled contaminations by E. Coli BL21, and two sterile batches for reference. Indirect analysis is indicated because the identification of contamination using Raman occurs when the mammalian cells are dead and is related to the lack of oxygen. The identification is not made when the contaminant cells are introduced but rather linked to other factors (e.g., cell death and no oxygen).

[0032] Raman data were collected during two sessions of one week each, with one acquisition every hour for each probe. FIG. 3 depicts a graph showing pre-processed Raman spectra for glucose, lipids, and proteins, according to embodiments of the disclosure. More specifically, FIG. 3 depicts a graph showing pre-processed Raman spectra for cell culture, the individual contribution of glucose, lipids, and proteins, can be identified according to embodiments of the disclosure. The presence of those contributions is a direct correlation with cell death. As shown, Raman shift in (cm'1) is shown on the x-axis and Raman intensity is plotted on the y-axis. Pre-processing algorithms are applied to theRaman spectra and highlight evolutions of the process and discriminate the contributions of its main components such as glucose at approximately between 1050 and 1200 cm’1, lipids at approximately between 1350 and 1500 cm’1, and proteins at approximately between 1560 and 1750 cm’1, which decrease during the run, as shown in Figure 3. A PCA is applied on these pre-processed spectra, i.e., Figure 4 shows PCA score plot for a batch of biological fluid with controlled contamination compared with a sterile batch for reference. At a dedicated time point, for example, after several days of culture, a contamination is induced in some culture and others are kept sterile as a reference. In FIG. 4, each point represents a Raman spectrum, for a given cell culture run, and new Raman spectra is taken every hour. FIG. 4 depicts Figures 4A and 4B which show comparisons between Raman spectra of contaminated and sterile cell cultures plotted on a PCA score plot, according to embodiments of the disclosure. When comparing the trend between sterile and contaminated cultures, it can be noted that the data from the contaminated batch start to deviate whereas the one from the sterile run keeps steady. This deviation is correlated to cell death, hence, it is an indirect correlation with the presence of bacteria when using a PCA score plot for this analysis. The “map,” as described earlier, is dedicated to qualitative analysis and clustering, here we compare trajectories, and identify clusters / differences. It is to be further understood that all data contemplated herein that is used for qualitative analysis or multivariate data analysis may comprise both sterile and contaminated samples.

[0033] Cellular death caused by culture contaminations was easily seen on Raman spectra analyzed with principal component analysis. However, contaminants were only detected in late-stage contamination. By way of example, the shift between early and late-stage contamination can be considered as the moment when there is an effect on the cell culture parameters, such as dissolved oxygen, decreases in viability or dramatic increases of glucose consumption. To proceed to quantitative-like detection, Partial Least Squares (PLS) model was evaluated to define a threshold of bacterial concentration and to warn the user of a potential contamination early in the process, as shown in Figure 5. FIG. 5 is a theoretical drawing, in this experiment. As shown, a sharp rise in bacterial concentration is evident, using the PLS model, soon after the introduction of the bacterial contamination. The bacterial concentration can be defined as a "culture parameter" to be monitored and modelized with a PLS model based on Raman spectra and offline values for bacterial concentration could be determined with theorical growth curves or with off-line optical density (OD) techniques. Different algorithms and machine learning techniques(clustering, neural networks, random forests, support vector machines) can be tested to detect contaminations in cell culture even earlier.

[0034] Detection of contamination by Raman probe seems to be even more efficient on stable media, typically in Cell Culture Media (CCM) preparation as described in first experiment or raw materials and for downstream processes (DSP) and for final product(s) for example measurements in a buffer.

[0035] Embodiments of the disclosure may prevent consequences due to culture contaminations. As the Raman measurements are made in real time, the system can detect not only contamination in early stage but also trigger an alarm to warn lab technicians of potential contaminations. This can help to identify more precisely the source of the contamination and so limit the risk for next batches, so saving costs and optimize the processes. Also, this can reduce the extension of bacteria and so cleaning processes are easier, cheaper, and more effective.

[0036] In-line, on-line, at-line and real-time measurement of Raman spectroscopy were performed by ProCellics™ Raman Analyzer associated with Bio4C® PAT Raman Software. The Raman spectra are collected, pre-processed and plotted using a PCA score plot. It has been operated on CCM preparation, in raw materials QC like described in the first experiment. Fed-batch CHO-cell cultures were carried out. Contaminations were intentionally generated using either BL21 strain of Escherichia coli or microorganisms from surrounding air. Raman measurements were taken every hour. Raman spectra were corrected from electronic sensor noise and cosmic rays. Raman spectra were pre-processed with dedicated algorithms (SNV, Savitzky-Golay derivative, spectral selection) and analyzed using multivariate data analysis (PCA, PLS). Various contaminants can be identified using the apparatus and methods described herein. For example, microorganisms that can be identified include, but are not limited to, bacteria, fungi, and yeast. It is further believed that viruses can also be identified.

[0037] Some methods according to the disclosure include detecting contamination in a biological fluid, comprising delivering an online sample of a biological fluid from a bioreactor during bioprocessing, taking Raman spectra data of the biological fluid, preprocessing the Raman spectra data; and analyzing the pre-processed data using qualitative mathematical tools or multivariate data analysis. A method to identify contaminants contemplated herein comprise collecting Raman spectra from both sterile and contaminated samples (clearly identified); collecting Raman spectra from new samples (without knowing if the samples are sterile or contaminated); pre-processing the spectra;and using a qualitative analytical tool to compare spectra to identify whether the unknown samples are contaminated. Pre-processing the spectra comprises, but is not limited to, analyzing the data using at least one of standard normal variate or normalization on a water band, a Savitzky-Golay derivative, and / or a spectral selection pre-processing. The Bio4C® PAT Raman software discussed herein comprises algorithms that remove baseline variations without altering the shape of spectral data. For example, the algorithm calculates a water band that is subsequently applied to spectral data of the biological fluid. More specifically, a water band may present Raman spectral data between 3100-3700 cm’1. This data is then applied on the whole spectra, typically over 0-4000 cm-1or specific Raman shift selection within this range.

[0038] At least one embodiment of the disclosure discloses a method for detecting contamination in a biological fluid. For example, a method starts. Thereafter, a step for collecting Raman spectra from a sterile sample is performed. Next, a step for collecting Raman spectra from a contaminated sample is performed. And then, a sample is collected, wherein Raman spectra from new samples (without knowing if the samples are sterile or contaminated) are obtained; pre-processing the spectra; and using a qualitative analytical tool to compare spectra to identify whether the unknown samples are contaminated. An alarm can then be triggered if the unknown sample is contaminated.

[0039] In a bioreactor, the Raman measurement can occur in the bioreactor or the sample can be delivered in another way to the Raman probe. In downstream processes, the Raman measurement can occur in other systems such as tanks, mixers, flow cells, etc., where the bioprocess step occurs, or in the fluidic path between bioprocess steps. At least one embodiment of the disclosure describes a method 100, as shown in FIG. 1, for detecting contamination in a fluid, comprising delivering an online sample of a biological fluid from a bioreactor or bioprocess system during bioprocessing at step 102; taking Raman spectra data of the biological fluid at step 104; pre-processing the Raman spectra data at step 106; and analyzing the pre-processed data using qualitative mathematical tools or multivariate data analysis at step 108. Thereafter, the method 100 ends. It should be noted that other steps can be added to the method 100. For example, the online sample may be drawn from a bioprocess using a sampler, e.g., an automated sampler, and, further, from an aseptic modular automated sampler. Furthermore, a microprocessor may be networked with the sampler, wherein a signal, indicating an out-of-specification condition, i.e., the presence of bacteria, fungi, and / or yeast within the bioreactor, is sent to a computer, tablet, smartphone, etc., so that a human is alerted, for example, a visual and / or an audio alarm, optionally atstep 110. Also, the signal may be adapted to take an automatic action within the bioreactor, e.g., add a processing aid, such as an anti-biotic, within the bioreactor through a port. The method 100 can be performed as incoming quality control for raw materials, during a bioprocessing run, such as a cell culturing process, or in processes downstream of the cell culturing step(s). It is to be noted that the method 100 can be performed whether sampled from a bioreactor bag, a delivery bag of raw materials, within a mixer bag, or other processes as are known to those in the art.

[0040] FIG. 6 displays graphical representations of four modes of Raman measurement, according to embodiments of the disclosure. Specifically, FIG. 6 illustrates the measurement of the biological fluid by a Raman probe 20 that can be done in-line (FIG. 6A), on-line (FIG. 6B), at-line (FIG. 6C), or off-line (FIG. 6D), wherein the Raman probe 20 is in immersion 22 or in non-contact 32 with the biological fluid, according to some embodiments of the disclosure. The Raman measurement can be done either on the biological fluid 11 inside a process bioreactor or biocontainer 10, or on the biological fluid 12 outside a process bioreactor or biocontainer 10.

[0001] As depicted, in FIG. 6 A, the Raman probe 20 can project into a bioreactor or biocontainer 10 in immersion mode 22 or, alternatively, the Raman probe 20 can acquire a measurement in a non-contact mode 32, i.e., the Raman probe 20 is external to the bioreactor or biocontainer 10. In this context, non-contact means that the Raman probe does not contact mechanically the biological fluid that is the object of the data capture or acquisition.

[0042] As depicted, as in FIG. 6B, an immersion mode Raman probe 20 can acquire a measurement in an outlet tubing 26 connected to a bioreactor or biocontainer 10 or within a recirculation loop 24 connected to a bioreactor or biocontainer 10. Similarly, a noncontact mode Raman probe 20 can acquire a measurement from an outlet tubing 26 connected to a bioreactor or biocontainer 10 or external to a recirculation loop 24 connected to a bioreactor or biocontainer 10.

[0043] As depicted, as in FIGS. 6C-6D, as with the other measurements steps, the biological fluid 11, 12 within the measurement cell or the vial 28 may be measured by Raman probe 20 either by immersion mode 22 or non-contact mode 32. FIGS. 6C-6D further show an interface / intermediate system 21 allowing non-contact, and / or non- immersive detection in which the Raman probe 20 is inserted in / connected to. A measurement cell or a vial 28, as shown in FIGS. 6C-6D, can be inserted in or be delivered in or in front of the interface / intermediate system 21 by an automated system, for example,the system of US Publ. No. 20220184599, assigned to EMD Millipore Corporation, Burlington, Mass, USA, which is incorporated by reference in its entirety or carried by a moving conveyor or robotic arm (not shown). Inversely, the probe can be moved in front of the measurement cell or a vial 28, for example carried by a moving conveyor or robotic arm (not shown). The measurement cell or the vial 28 may be centrifugated or mixed / shaken to respectively concentrate or put in suspension the contamination particles. The interface / intermediate system 21 described above may include a thermal control feature.

[0044] The system 200 may be connected to a computer or a network (not shown) for automated measurement when the measurement cell or the vial 28 is ready to be measured. Multiple measurement of a same sample can be done to either have a reproducibility measurement or analyze and quantify the evolution of contamination through a certain amount of time (e.g., over one, five, or ten minutes or longer).

[0045] The Raman probe 20 is connected to a Raman analyzer 41, which is exterior to the bioreactor or biocontainer 10, using a conduit 42. The Raman analyzer 41 may include a laser 43, which generates a laser beam that travels through a conduit 42 to the Raman probe 20. This conduit 42 may be a fiber optic cable. The Raman analyzer 41 may include an optical detector 44, which may include a charge-coupled device (CCD) or photodetector. The Raman analyzer 41 also includes a processing unit 45 to control the laser 43 and interpret the output from the optical detector 44.

[0046] A controller 50 may be in communication with the Raman analyzer 41. The controller 50 may include a processing unit and an associated memory device. The processing unit may be any suitable component, such as a microprocessor, embedded processor, an application specific circuit, a programmable circuit, a microcontroller, or another similar device. This memory device contains the instructions, which, when executed by the processing unit, enable the controller 50 to perform the functions described herein. The controller 50 may receive data from the Raman analyzer 41 and provide commands or instructions to the Raman analyzer 41. In the embodiment shown in FIG. 6B, the controller 50 is separate from the Raman analyzer 41. However, in other embodiments, these two components may be combined. In other embodiments, the controller 50 may transmit processed data to the bioreactor controller (not shown). It is to be understood that the controller 50 can be incorporated within any of the embodiments disclosed herein, for example, any of the embodiments in FIGS. 6A-6D.

[0047] In some embodiment of the disclosure, the method can consist in measuring at- line the Raman spectra of a sample through a vial, pre-processing spectral data and analyzing the preprocessed data using clustering algorithm such as Partial Least Square Discriminant Analysis to determine the class label of the sample; for example, the class label of the sample can be “Sterile” or “Contaminated”. In a first step, a model is created by measuring a certain number of samples with a known class label (“Sterile” or “Contaminated”). In order to be robust, the model should be trained with a sufficient number of samples from the two classes. Each sample data is projected on the component axis defined by the PLS-DA which maximizes the separation of the data in two groups, each group corresponding to a class. Then the model can be used for classification of new samples with unknown class label. The projection of the sample in either the “Sterile” (Not contaminated) or “Contaminated” area of the model defines its classification.

[0048] An experiment was conducted in three milliliter vials of bioprocess fluid with five contaminated samples and five sterile samples. The contamination of the fluid is simulated by the presence of mammalian cells. The sterility of the fluid is simulated by the absence of mammalian cells. A model was built beforehand with 10 calibration samples (five contaminated samples and five sterile samples). Then ten unknown test samples were measured five times over a period of 200 seconds.

[0049] FIG. 7 depicts a graph plotting a classification of data using a PLS-DA clustering algorithm, according to embodiments of the disclosure. FIG. 7A shows the model training phase of the PLS-DA algorithm based on the five known contaminated samples and five known sterile samples. FIG. 7B shows the monitoring phase with all five measurements of the ten test samples analyzed with the PLS-DA model, giving access to the classification between test samples with cells (“Contaminated”) and test samples without cells (“Not Contaminated”). The analysis with PLS-DA automatically highlights differences between contaminated biological fluid and non-contaminated biological fluid. It provides an interpretable result based on a threshold, or a limit or a separation between 2 classes and facilitate the automation of decision, for example to stop the process or to trigger an alarm.

[0050] All ranges for formulations recited herein include ranges therebetween and can be inclusive or exclusive of the endpoints. Optional included ranges are from integer values therebetween (or inclusive of one original endpoint), at the order of magnitude recited or the next smaller order of magnitude. For example, if the lower range value is 0.2, optional included endpoints can be 0.3, 0.4, . . . 1.1, 1.2, and the like, as well as 1, 2, 3 and the like; if the higher range is 8, optional included endpoints can be 7, 6, and the like, as well as 7.9,7.8, and the like. One-sided boundaries, such as 3 or more, similarly include consistent boundaries (or ranges) starting at integer values at the recited order of magnitude or one lower. For example, 3 or more includes 4, or 3.1 or more.

[0051] Reference throughout this specification to “one embodiment,” “certain embodiments,” “one or more embodiments,” “some embodiments,” or “an embodiment” indicates that a feature, structure, material, or characteristic described in connection with the embodiment is included in at least one embodiment of the disclosure. Therefore, the appearances of the phrases such as “in one or more embodiments,” “in certain embodiments,” “in one embodiment,” “some embodiments,” or “in an embodiment” throughout this specification are not necessarily referring to the same embodiment.

[0052] Although some embodiments have been discussed above, other implementations and applications are also within the scope of the following claims. Although the specification describes, with reference to particular embodiments, it is to be understood that these embodiments are merely illustrative of the principles and applications of the present disclosure. It is therefore to be further understood that numerous modifications may be made to the illustrative embodiments and that other arrangements and patterns may be devised without departing from the spirit and scope of the embodiments according to the disclosure. Furthermore, particular features, structures, materials, or characteristics may be combined in any suitable manner in any one or more of the embodiments.

[0053] Publications of patent applications and patents and non-patent references cited in this specification are herein incorporated by reference in their entirety in the entire portion cited as if each individual publication or reference were specifically and individually indicated to be incorporated by reference herein as being fully set forth. Any patent application to which this application claims priority is also incorporated by reference herein in the manner described above for publications and references.

Claims

CLAIMS1. A method for detecting contamination, testing sterility of a final product, or measuring bioburden in a biological fluid, comprising: delivering a sample of a biological fluid from a container, including biocontainer, a vessel, a mixer bag, a delivery bag, a measurement cell, or a vial; taking Raman spectra data of the biological fluid; pre-processing the Raman spectra data; and analyzing the pre-processed data using clustering algorithms or qualitative mathematical tools or multivariate data analysis.

2. The method of claim 1, wherein the pre-processing of the data is at least one of standard normal variate or normalization on a water band, Savitzky-Golay derivative, and spectral selection pre-processing.

3. The method of claims 1 or 2, wherein the qualitative mathematical tool is principal component analysis.

4. The method of claims 1-2, wherein the multivariate analysis or clustering algorithm is partial least squares analysis or partial least squares discriminant analysis, respectively.

5. The method of claims 1-4, wherein the biological fluid is a cell culture medium or buffer or final product.

6. The method of claims 1-5, wherein the contamination is bacteria, fungi or yeast.

7. The method of claims 1-6, wherein the sample is an online, in-line, off-line or at- line sample.

8. The method of claims 1-7, wherein the pre-processing of the Raman spectral data is calculating a water band that is subsequently applied to spectral data of the biological fluid.

9. The method of claim 8, wherein the water band is calculated over a Raman spectral band between 3100-3700 (cm'1), and subsequently applied on the whole spectra, typically over 0-4000 cm-1or specific Raman shift selection within this range.

10. The method of claims 1-9, wherein an alarm is triggered at the onset of detection of a contamination.

11. The method of claims 1-10, wherein the method is performed during a bioprocessing run or a quality control of incoming raw materials or sterility testing of a final product.

12. The method of claim 11, wherein during the bioprocessing run, a sample is measured in-line or at-line.

13. The method of claims 1-12 wherein the pre-processed data used for qualitative analysis or multivariate data analysis may comprise both sterile and contaminated samples.

14. The method of claims 1-12, wherein the measurement is performed in immersion or in non-contact with the biological fluid.

15. The method of claim 1, wherein the sample is measured during a downstream process.

Images