Virus analysis method using Raman spectroscopy

Abstract

This invention provides a method for monitoring and evaluating viral titer and / or viral component abundance. [Solution] A method for determining the ratio of viral nucleic acid to a virus containing one or more viral structural molecules in a viral culture using Raman spectroscopy, comprising the steps of providing a viral culture, irradiating the viral culture with a light source, and measuring the total intensity of Raman scattered light in each of a first plurality of wavenumber ranges to obtain a first wavenumber intensity dataset of the viral culture, wherein the steps of irradiating the viral culture with a light source and measuring the total intensity of Raman scattered light are performed directly (in situ) on the culture medium of the viral culture.

Description

[Technical Field]

[0001] Field of Invention This invention relates to the use of Raman spectroscopy for monitoring and evaluating viral titers and / or viral component abundances. [Background technology]

[0002] Background of the Invention The production of viral vectors is a critical process step for the production of many cells and gene therapies, and the growth of this industry is driving increased demand for viral vector supplies. Given this, it is crucial that analytical tools are available to ensure the monitoring and optimization of the viral vector production process, as well as the quantification and characterization of viral vector products. Both the demand for viral vectors and the production costs place particular importance on achieving good viral vector titers during production. Currently, the measurement of physical viral titers is typically performed using standard analytical techniques; for example, ELISA for evaluating p24 or qPCR for measuring viral RNA in the case of lentiviral titers, and RT-qPCR using ITR-targeting primers in the case of AAVs are frequently used. However, these methods are time-consuming, often inaccurate, and require culture medium sampling. Therefore, these methods only provide retrospective measurements of viral concentrations. Thus, a new method for measuring viral titers is needed.

[0003] Furthermore, for efficient virus or viral particle production processes, particularly those used to produce viral particles for medical applications, it is advantageous to monitor the ratio of viral nucleic acid to viruses containing one or more viral structural molecules to maximize the proportion of functional viral particles produced. Non-functional viral particles, such as hollow particles, are generally considered waste in the production process and, if administered to patients, can cause problems such as undesirable immune responses. Currently, the most accurate method for quantifying the ratio of viral nucleic acid to viruses containing one or more viral structural molecules is transmission electron microscopy (TEM). Another method is to perform ELISA and qPCR experiments to calculate the amounts of viral protein and nucleic acid, respectively. However, these quantification methods of TEM, ELISA, and qPCR can only be performed retrospectively. Therefore, there is currently no method to quantify the abundance of viral components in real time in order to calculate the proportion of functional viral particles in a sample.

[0004] Raman spectroscopy is a vibrational spectroscopy method that has been shown to be particularly useful for process analysis techniques (PAT) applications where molecular information is required. This technique is based on detecting wavelength shifts of photons inelastically scattered by molecules present in a sample (such differences in photon wavelengths relate to energy lost from the photon by changing the vibrational state of a particular molecule from the ground state to a first excited state, or to energy gained by photons from molecules undergoing a downward transition from an excited vibrational state to the ground state). This technique offers many advantages over previously used methodologies in PAT applications, particularly because the signal generated by water is relatively weak compared to other systems, facilitating bioprocess monitoring, cell culture analysis, and protein analysis in solution. Raman spectroscopy is used, for example, in dispersion testing and to determine the identity of raw materials and cell culture media; to characterize polymer products; to analyze drug formulations, batch-to-batch variability, contamination, media degradation, cell density including viable cell density and total cell density, protein structure, and protein stability; for polymer and fiber analysis; for material ID testing; and for glucose, glutamine, lactate, and other proteins. It has been used for the quantitative determination of salts and ammonia. Buckley and Ryder (2017) provide a review of the applicability of Raman spectroscopy.

[0005] Conventional Raman spectroscopy may be associated with the generation of weak signals, which has been reported to affect its use in highly sensitive quantitative analysis often required in the biological field. Indeed, the concentration limit for glucose detection using conventional Raman spectroscopy has been reported to be 0.6 mM, and the concentration limit for phenylalanine is 1.1 mM (estimated to be equivalent to 0.11 mg / ml and 0.18 mg / ml, respectively) (Buckley and Ryder, 2017, Applied Spectroscopy, (71, pp. 1085-1116). Considering this, conventional Raman spectroscopy has primarily been used in the art to evaluate cell culture media and cell proliferation. Other methods have been employed when more sensitive measurements are needed to detect entities below the reported detection limits of conventional Raman spectroscopy, such as viral particles. In this regard, Lee et al (2015) describe the use of surface-enhanced Raman spectroscopy (SERS) for the detection of HIV-1. Unlike conventional Raman, SERS enhances signals from molecules attached to nanoscale roughened metal surfaces such as Ag, Cu, or Au. Lee et al used an indium tin oxide substrate fabricated with Au nanodots containing a bound anti-gp120 antibody fragment for the specific binding of HIV-1 virus-like particles to ensure the generation of an enhanced signal for HIV-1 virus-like particle detection. However, SERS may have limitations, such as the need to pre-prepare the substrate with appropriate immuno-interaction molecules, which limits the general application of the technology and increases the associated costs. Furthermore, due to the nature of its requirement to bind to the target entity, SERS is always invasive within the sample.

[0006] Therefore, further methods are needed to characterize and evaluate entities that may be associated with sensitive or weak signals, such as viruses, and to evaluate such entities in situ. Additionally, methods are needed to monitor viral component abundance in real time to monitor and maximize the proportion of functional viral particles produced. [Overview of the project]

[0007] Surprisingly, contrary to teachings in the art, we have shown that viral titers can be accurately and sensitively quantified in real time using conventional Raman spectroscopy. This can be achieved by directly irradiating the viral culture medium with a light source and acquiring Raman spectral data, as further described herein. Thus, this method eliminates the need for processing the viral culture medium. We have shown that viral titers determined by conventional Raman spectroscopy according to the method herein are comparable to offline titer measurements (such as those measured by offline assays such as RT-qPCR and p24 ELISA), and that conventional Raman spectroscopy used according to the method herein provides a rapid and reliable alternative method for evaluating viral titers. As mentioned above, this finding is particularly unexpected given the prior art, which has been understood to be associated with the generation of weak signals, and that when applied to bulk solution systems (especially systems with large scattering variations and unwanted fluorescence background signals), low concentrations of virus in solution fall below the typical detection limit by conventional Raman spectroscopy. This finding is also particularly unexpected, considering prior art that requires processing the viral culture medium to enable the use of Raman spectroscopy (e.g., processing the viral culture medium to concentrate the virus before analysis using Raman spectroscopy, or to prevent biofilm formation).

[0008] Furthermore, the inventors used conventional Raman spectroscopy to determine the amount of viral components in real time. We demonstrated that monitoring can be performed in real time, particularly quantifying the amount of viral nucleic acid and viral structural molecules present. Therefore, using Raman spectroscopy, it is possible to determine the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in a sample, and thus determine the proportion of functional viral particles produced.

[0009] The inventors identified a set of spectral variables crucial for enabling predictions in a model that processes real-time Raman spectroscopy data, in order to achieve real-time measurement of viral titer and / or viral nucleic acid and viral structural molecule abundances. The inventors established a stratified range of an increasing number of variables with different importance thresholds to provide a variable range for accurate prediction of viral titer and / or viral nucleic acid and viral structural molecule abundances when using Raman spectroscopy. The inventors demonstrated that Raman spectroscopy can be used to identify the initiation, production stages, and termination of the viral production process.

[0010] Therefore, the present invention relates to the use of Raman spectroscopy for monitoring and / or evaluating the abundance of viral nucleic acid and viral structural molecules in a sample. In other words, the present invention relates to a method for monitoring and / or evaluating the abundance of viral nucleic acid and viral structural molecules in a sample, comprising the step of analyzing the Raman spectrum of the virus-containing sample using a multivariate model to determine the abundance of viral nucleic acid and viral structural molecules. Such a method may further include the step of performing Raman spectroscopy on the sample (i.e., obtaining a spectrum).

[0011] The present invention more specifically relates to the use of Raman spectroscopy for monitoring and / or evaluating the initiation, production stages, and / or termination of the virus production process. As described above, the inventors have identified that Raman spectroscopy can be used to determine the abundance of viral nucleic acid and viral structural molecules, and in particular, the inventors have identified specific wavenumber ranges that need to be evaluated for this purpose. The intensity of such peaks can be determined by the method of the present invention, and the intensity of such peaks may result in the generation of a fingerprint that can be evaluated in a multivariate model to determine the abundance of viral nucleic acid and viral structural molecules. In a preferred embodiment of the present invention, the abundance of viral nucleic acid and viral structural molecules to be monitored and / or evaluated is the abundance of viral nucleic acid and viral structural molecules of adeno-associated virus.

[0012] One advantage of the present invention is that viral nucleic acid abundance, viral structural molecule abundance, and the ratio of viral nucleic acid to viruses comprising one or more viral structural molecules can be continuously monitored in real time. There is no need to process a sample from the virus culture medium to estimate the viral nucleic acid abundance, the viral structural molecule abundance, and the ratio of viral nucleic acid to viruses comprising one or more viral structural molecules. Measurements can be performed in situ as needed. In other words, measurements can be performed directly on the virus culture medium in a growth incubator. Measurements can be performed ex situ as needed. In other words, measurements can be performed directly on the virus culture medium in an aliquot of virus culture medium collected from the growth incubator or separated from the main chamber of the growth incubator. Regardless of whether the measurement is performed in situ or ex situ, the measurement can be performed directly on the virus culture medium without the need for further processing of the virus culture medium. This type of approach is sometimes referred to as "in-line" or "at-line" analysis. Accordingly, more accurate trends of viral nucleic acid abundance and viral structural molecule abundance can be generated without the risk of contamination. Therefore, the method of the present invention is faster and simpler than conventional offline methods that require processing of the virus culture medium. In particular, since the production phase of culture can be measured much more accurately, the timing of the end / harvest phase of virus production becomes more accurate , allowing the process to be stopped at an appropriate time, which may potentially reduce the cost of the production process.

[0013] Accordingly, the present invention provides a method for determining the ratio of viral nucleic acid to viruses comprising one or more viral structural molecules in a sample using Raman spectroscopy, the method comprising: (a) providing a sample and irradiating the sample with a light source; (b) (i) measuring the total intensity of Raman scattered light within each one of a first plurality of wavenumber ranges, respectively, to obtain a first wavenumber intensity data set of a sample, wherein the first plurality of wavenumber ranges are preselected and characteristic of viral nucleic acid in the sample; (ii) performing a first set of mathematical data processing steps on the first wavenumber intensity data set; (iii) determining the viral nucleic acid content of the sample based on the output of said first set of mathematical data processing steps; (c) (i) measuring the total intensity of Raman scattered light within each one of a second plurality of wavenumber ranges, respectively, to obtain a second wavenumber intensity data set of a sample, wherein the second plurality of wavenumber ranges are preselected and characteristic of one or more viral structural molecules of a virus in the sample; (ii) performing a second set of mathematical data processing steps on the second wavenumber intensity data set; (iii) determining a virus content comprising one or more viral structural molecules in the sample based on the output of said second set of mathematical data processing steps; (d) determining a ratio of viral nucleic acid to virus comprising one or more viral structural molecules in the sample based on the values determined in step (b)(iii) and step (c)(iii).

[0014] In the method defined above, steps (b) and (c) can be performed in the order of (b) followed by (c), or in the order of (c) followed by (b). The exact order in which the steps are performed is not essential, provided that the values identified in steps (b)(iii) and (c)(iii) are determined such that the ratio can be determined in step (d). Furthermore, steps (b) and (c) can be performed simultaneously, since mathematical data processing may enable the first and second wavenumber intensity data sets to be processed concurrently to provide the values identified in steps (b)(iii) and (c)(iii).

[0015] In the method defined above, the step of performing the first and second sets of mathematical data processing steps on the first and second wavenumber intensity datasets is: (i) optionally, a step of normalizing the wavenumber signal intensity data, which is performed by preprocessing the signal intensity data using one or more preprocessing analysis methods such as first derivative, second derivative, standard normalized variable (SNV) method, polynomial fitting, multi-polynomial fitting, Morifier method, piecewise polynomial fitting (PPF) method, or adaptive iteratively reweighted penalty least squares (airPLS) method. (ii) A step of obtaining model parameters by applying a multivariate regression algorithm, such as a partial least squares (PLS) regression algorithm, to wavenumber signal intensity data, wherein the PLS algorithm is optionally a nonlinear iterative partial least squares (NIPALS) regression algorithm or a neural network, (iii) Using model parameters obtained by applying a multivariate regression algorithm to the signal intensity data, determine the viral nucleic acid content of the sample. The procedure may include a step of determining the virus content, which includes one or more viral structural molecules.

[0016] In any of the methods defined above, the light source used to irradiate the sample can be a laser, and the sample can be irradiated with light having a wavelength of 785 nm.

[0017] In any of the methods defined above, Raman scattered light can be detected using a charge-coupled device (CCD).

[0018] In one embodiment of the method defined above, the first set of wavenumber ranges of the Raman spectrum measured to obtain a first wavenumber intensity dataset of a sample may include four or more of the wavenumber ranges 1 to 12 listed in Table 1, where VIP is 1.00 or greater, or the set of wavenumber ranges of the measured Raman spectrum may include six or more of the wavenumber ranges 1 to 12 listed in Table 1, where VIP is 1.00 or greater, or the set of wavenumber ranges of the measured Raman spectrum may include eight or more of the wavenumber ranges 1 to 12 listed in Table 1, where VIP is 1.00 or greater, or the set of wavenumber ranges of the measured Raman spectrum may include ten or more of the wavenumber ranges 1 to 12 listed in Table 1, where VIP is 1.00 or greater, or the set of wavenumber ranges of the measured Raman spectrum may include all twelve of the wavenumber ranges 1 to 12 listed in Table 1, where VIP is 1.00 or greater. In any of these methods, the virus can preferably be an adeno-associated virus (AAV).

[0019] Alternatively, the first set of wavenumber ranges of the measured Raman spectrum may include four or more of the wavenumber ranges 13–22 with a VIP of 1.25 or greater, as listed in Table 1; or the set of wavenumber ranges of the measured Raman spectrum may include six or more of the wavenumber ranges 13–22 with a VIP of 1.25 or greater, as listed in Table 1; or the set of wavenumber ranges of the measured Raman spectrum may include eight or more of the wavenumber ranges 13–22 with a VIP of 1.25 or greater, as listed in Table 1; or the set of wavenumber ranges of the measured Raman spectrum may include all ten of the wavenumber ranges 13–22 with a VIP of 1.25 or greater, as listed in Table 1. In any of these methods, the virus may preferably be an adeno-associated virus (AAV).

[0020] Alternatively, the first set of wavenumber ranges of the measured Raman spectrum may include four or more wavenumber ranges 23–30 with a VIP of 1.50 or greater, as listed in Table 1; or the set of wavenumber ranges of the measured Raman spectrum may include six or more wavenumber ranges 23–30 with a VIP of 1.50 or greater, as listed in Table 1; or the set of wavenumber ranges of the measured Raman spectrum may include all eight wavenumber ranges 23–30 with a VIP of 1.50 or greater, as listed in Table 1. In any of these methods, the virus may preferably be an adeno-associated virus (AAV).

[0021] In the same or another embodiment of the method defined above, the second set of wavenumber ranges of the Raman spectrum measured to obtain a second wavenumber intensity dataset of the sample may include four or more wavenumber ranges 1 to 20 with a VIP of 1.00 or greater, as listed in Table 2, or the set of wavenumber ranges of the measured Raman spectrum may include six or more wavenumber ranges 1 to 20 with a VIP of 1.00 or greater, as listed in Table 2, or the set of wavenumber ranges of the measured Raman spectrum may include The measured Raman spectrum may include eight or more wavenumber ranges from 1 to 20 with a VIP of 1.00 or greater, as listed in Table 2; or the measured Raman spectrum may include ten or more wavenumber ranges from 1 to 20 with a VIP of 1.00 or greater, as listed in Table 2; or the measured Raman spectrum may include twelve or more wavenumber ranges from 1 to 20 with a VIP of 1.00 or greater, as listed in Table 2; or the measured Raman spectrum may include all twenty wavenumber ranges from 1 to 20 with a VIP of 1.00 or greater, as listed in Table 2. In any of these methods, the virus may preferably be an adeno-associated virus (AAV).

[0022] Alternatively, the second set of wavenumber ranges of the measured Raman spectrum may include four or more of the wavenumber ranges 21–33 listed in Table 2, where VIP is 1.25 or greater; or the set of wavenumber ranges of the measured Raman spectrum may include six or more of the wavenumber ranges 21–33 listed in Table 2, where VIP is 1.25 or greater; or the set of wavenumber ranges of the measured Raman spectrum may include eight or more of the wavenumber ranges 21–33 listed in Table 2, where VIP is 1.25 or greater; or the set of wavenumber ranges of the measured Raman spectrum may include ten or more, eleven or more, or twelve of the wavenumber ranges 21–33 listed in Table 2, where VIP is 1.25 or greater; or the set of wavenumber ranges of the measured Raman spectrum may include all thirteen of the wavenumber ranges 21–33 listed in Table 2, where VIP is 1.25 or greater. In any of these methods, the virus can preferably be an adeno-associated virus (AAV).

[0023] Alternatively, the second set of wavenumber ranges of the measured Raman spectrum may include four or more of the wavenumber ranges 34–40 with a VIP of 1.50 or greater, as listed in Table 2; or the set of wavenumber ranges of the measured Raman spectrum may include five or six of the wavenumber ranges 34–40 with a VIP of 1.50 or greater, as listed in Table 2; or the set of wavenumber ranges of the measured Raman spectrum may include all seven of the wavenumber ranges 34–40 with a VIP of 1.50 or greater, as listed in Table 2. In any of these methods, the virus may preferably be adeno-associated virus (AAV).

[0024] In another embodiment of the method defined above, the first set of wavenumber ranges of the Raman spectrum measured to obtain a first wavenumber intensity dataset of a sample may include five or more wavenumber ranges 1 to 28 with a Variable Importance Index (VIP) of 1.00 or greater, as listed in Table 3, or the set of wavenumber ranges of the measured Raman spectrum may include ten or more wavenumber ranges 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3, or the set of wavenumber ranges of the measured Raman spectrum may include wavenumber ranges 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3. The wavenumber ranges of the measured Raman spectrum may include 15 or more of the ranges 1 to 28, or 20 or more of the wavenumber ranges 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3, or 25 or more of the wavenumber ranges 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3, or 28 of the wavenumber ranges 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3. In any of these methods, the virus may preferably be a lentivirus.

[0025] Alternatively, the first multiple wavenumber ranges of the measured Raman spectrum are listed in Table 3. The Raman spectrum can include five or more wavenumber ranges from 29 to 59 with a Variable Importance Index (VIP) of 1.25 or higher, or the multiple wavenumber ranges of the measured Raman spectrum can include ten or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3, or the multiple wavenumber ranges of the measured Raman spectrum can include fifteen or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3, or the multiple wavenumber ranges of the measured Raman spectrum can include five or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3. The wavenumber ranges may include 20 or more of the wavenumber ranges 29 to 59, or the multiple wavenumber ranges of the measured Raman spectrum may include 25 or more of the wavenumber ranges 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3, or the multiple wavenumber ranges of the measured Raman spectrum may include 30 of the wavenumber ranges 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3, or the multiple wavenumber ranges of the measured Raman spectrum may include all 31 of the wavenumber ranges 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3. In any of these methods, the virus may preferably be a lentivirus.

[0026] Alternatively, the first set of wavenumber ranges of the measured Raman spectrum may include five or more of the wavenumber ranges 60–81 with a Variable Importance Index (VIP) of 1.50 or greater, as listed in Table 3; or the set of wavenumber ranges of the measured Raman spectrum may include ten or more of the wavenumber ranges 60–81 with a VIP of 1.50 or greater, as listed in Table 3; or the set of wavenumber ranges of the measured Raman spectrum may include fifteen or more of the wavenumber ranges 60–81 with a VIP of 1.50 or greater, as listed in Table 3; or the set of wavenumber ranges of the measured Raman spectrum may include twenty or more of the wavenumber ranges 60–81 with a VIP of 1.50 or greater, as listed in Table 3; or the set of wavenumber ranges of the measured Raman spectrum may include all 22 of the wavenumber ranges 60–81 with a VIP of 1.50 or greater, as listed in Table 3. In any of these methods, the virus can preferably be a lentivirus.

[0027] In any of the methods defined above, the nucleic acid may include either a viral DNA genome or a viral RNA genome.

[0028] In any of the methods defined above, one or more viral structural molecules may include one or more viral proteins (e.g., one or more nucleoproteins and / or one or more capsomeres), one or more viral glycans, one or more glycosylated viral molecules (e.g., glycosylated viral proteins), and / or one or more viral lipids.

[0029] In any of the methods defined above, the ratio can provide a measure of functional viral titer.

[0030] In any of the methods defined above, the sample can be a viral culture. The viral culture may be contained in a bioreactor. In any of these methods, the step of irradiating the viral culture with a light source and measuring the total intensity of the Raman scattered light can be performed directly on the culture medium of the viral culture (in situ). Alternatively, the step of irradiating the viral culture with a light source and measuring the total intensity of the Raman scattered light can be performed directly on an aliquot of the culture medium taken from the viral culture (ex situ).

[0031] Any of the methods defined above involves a first step of determining the ratio of viral nucleic acid to a virus containing one or more viral structural molecules at an initial point in time, and a second step of determining one or more at a later point in time. The method may include one or more further steps of determining the ratio of viral nucleic acid to the viral structural molecules described above, and further including a step of measuring the change in the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in a sample between time points, each step being performed by one of the methods defined above, preferably each step being performed by the same method. In any of such methods, the ratio of viral nucleic acid to viruses containing one or more viral structural molecules can be repeatedly determined over a period of time to provide a measure of the change in the ratio in real time. The change in the ratio in the sample can be used to determine the initiation stage, production stage, and / or steady stage of the viral production process. Any of such methods can be used to determine the optimal conditions for the viral production process. Any of such methods can be used to evaluate downstream processes of the viral production process.

[0032] Any of the methods defined above may include a step of comparing the ratio obtained thereby with the ratio obtained from the same sample by an alternative method, which may optionally be qPCR, RT-qPCR, ELISA, or visual determination by transmission electron microscopy.

[0033] The present invention also provides a method for determining the degree of viral infection in an individual using Raman spectroscopy, comprising the step of determining the ratio of viral nucleic acid to virus containing one or more viral structural molecules in a sample by performing any one of the methods defined above, wherein the sample is a sample previously obtained from the individual. The sample may be a blood, saliva, sputum, plasma, serum, cerebrospinal fluid, urine, or fecal sample. In any of such methods, the ratio in the sample from the subject may be compared to one or more ratio measurements previously obtained for the individual's infection to provide an indication of the individual's stage of infection.

[0034] The present invention further provides a method for determining the ratio of viral nucleic acid to a virus containing one or more viral structural molecules in a sample using Raman spectroscopy, (a) (i) Providing a first wavenumber intensity dataset of a sample, the first dataset being obtained by irradiating the sample with a light source and measuring the total intensity of Raman scattered light in each of a first set of wavenumber ranges, wherein the first set of wavenumber ranges in the Raman spectrum are selected as characteristic of the viral nucleic acid in the sample. (ii) The step of performing a first set of mathematical data processing steps on a first wavenumber intensity dataset, (iii) A step of determining the nucleic acid content of the sample based on the output of the first set of mathematical data processing steps, (b) (i) Providing a second wavenumber intensity dataset of a sample, the second dataset being obtained by irradiating the sample with a light source and measuring the total intensity of Raman scattered light in each of a second set of wavenumber ranges, wherein the second set of wavenumber ranges in the Raman spectrum are selected as characteristic of one or more viral structural molecules of the virus in the sample. (ii) The step of performing a second set of mathematical data processing steps on a second wavenumber intensity dataset, (iii) A step of determining the virus content in the sample, which includes one or more viral structural molecules, based on the output of the second set of mathematical data processing steps, (c) Based on the values ​​determined in steps (a)(iii) and (b)(iii), The present invention provides a method comprising the step of determining the ratio of viral nucleic acid to a virus containing one or more viral structural molecules in a sample.

[0035] The method described above can be carried out according to the steps defined in any one of the methods described and defined herein.

[0036] In the method described above, steps (a) and (b) can be performed in the order of (a) followed by (b), or in the order of (b) followed by (a). The exact order in which the steps are performed is not essential if the values ​​identified in steps (a)(iii) and (b)(iii) are determined in such a way that the ratio can be determined in step (c). Furthermore, steps (a) and (b) can be performed simultaneously, as mathematical data processing may enable simultaneous processing of the first and second wavenumber intensity datasets to provide the values ​​identified in steps (a)(iii) and (b)(iii).

[0037] The present invention further provides the use of Raman spectroscopy to determine the ratio of viral nucleic acids to viruses containing one or more viral structural molecules in a sample. In any such use, the ratio can be determined based on measurements of the intensity of Raman scattered light obtained from the sample after irradiation of the sample with a light source, the intensity of which is measured from a first set of wavenumber ranges of the Raman spectrum characteristic of the viral nucleic acids in the sample and from a second set of wavenumber ranges of the Raman spectrum characteristic of one or more viral structural molecules of the viruses in the sample.

[0038] In any such use, the sample may be a viral culture, and optionally, the viral culture may be contained in a bioreactor. The step of measuring the total intensity of Raman scattering light can be performed directly (in situ) on the medium of the viral culture. Alternatively, the step of measuring the total intensity of Raman scattering light can be performed directly (ex situ) on an aliquot of the medium taken from the viral culture.

[0039] In any such use, the proportion in the sample can be determined at the initial point in time and at one or more subsequent points in time, and the change in the proportion in the sample between time points is calculated. In any such use, the proportion in the sample is repeatedly quantified to provide a measure of the change in the proportion in real time. In any such use, the viral titer in the sample can be quantified by performing any of the methods described and defined herein.

[0040] In any of the methods or uses defined above, the virus in the sample does not have to be HIV-1 or HIV-1 virus-like particles (HIV-1 VLPs).

[0041] In any of the methods or uses defined above, the Raman spectroscopy does not have to be surface-enhanced Raman spectroscopy.

[0042] The present invention further provides a method for constructing a multivariate data processing model that can determine the viral content, including one or more viral structural molecules, in a sample from a wavenumber intensity dataset obtained from Raman spectroscopy for the sample, (a) The steps include providing a sample and irradiating the sample with a light source, (b) A step of obtaining a wavenumber intensity dataset of a sample by measuring the total intensity of Raman scattered light in each of a plurality of wavenumber ranges, wherein the plurality of wavenumber ranges are pre-selected and characteristic of one or more viral structural molecules of the virus in the sample, (c) The step includes obtaining normalized wavenumber signal intensity data, which is a step in which This is performed by preprocessing signal intensity data using preprocessing analysis methods such as first-order differential calculus, second-order differential calculus, standard normal variable (SNV) method, polynomial fitting method, multi-polynomial fitting method, Morifier method, piecewise polynomial fitting (PPF) method, or adaptive iteratively reweighted penalty least squares (airPLS) method. (d) A step of obtaining model parameters by applying a multivariate regression algorithm, such as a partial least squares (PLS) regression algorithm, to preprocessed signal intensity data, wherein the PLS algorithm is optionally a nonlinear iterative partial least squares (NIPALS) regression algorithm or a neural network, and calibration is performed, and the preprocessed signal intensity data is compared with viral titer data obtained under the same sample conditions using non-Raman spectroscopy such as qPCR, RT-qPCR, ELISA, or by visual determination by transmission electron microscopy, (e) The step of inferring response values ​​using model parameters obtained from preprocessed data, (f) Includes the steps of variable selection, optionally performing a Variable Importance Index (VIP) to identify Raman spectral variables, (g) Optionally, perform one or more further rounds of modeling by reapplying steps (d) to (f), including a step to remove non-essential variables. This invention provides a method for determining the viral content of a sample, including one or more viral structural molecules, using model parameters obtained for identified Raman spectral variables derived from a multivariate data processing model.

[0043] The present invention also further provides a method for constructing one or more multivariate data processing models that can determine the ratio of viral nucleic acids to viruses containing one or more viral structural molecules in a sample from a wavenumber intensity dataset obtained from Raman spectroscopy for a sample, (a) The steps include providing a sample and irradiating the sample with a light source, (b) (i) a step of obtaining a first wavenumber intensity dataset of a sample by measuring the total intensity of Raman scattered light in each of a first plurality of wavenumber ranges, wherein the first plurality of wavenumber ranges are pre-selected and characteristic of viral nucleic acids in the sample. (ii) A step of obtaining a second wavenumber intensity dataset of a sample by measuring the total intensity of Raman scattered light in each of a second plurality of wavenumber ranges, wherein the first plurality of wavenumber ranges are pre-selected and characteristic of one or more viral structural molecules of the virus in the sample. (c) The step includes obtaining normalized wavenumber signal intensity data from the first and second wavenumber intensity datasets, which is performed by preprocessing the signal intensity data using a preprocessing analysis method such as first derivative, second derivative, standard normalized variable (SNV) method, polynomial fitting, multi-polynomial fitting, Morifier method, piecewise polynomial fitting (PPF) method, or adaptive iteratively reweighted penalty least squares (airPLS) method. (d) A step of obtaining model parameters to be applied to first and second wavenumber intensity datasets by applying a multivariate regression algorithm, such as a partial least squares (PLS) regression algorithm, to one of each of the preprocessed signal intensity datasets, wherein optionally the PLS algorithm is a nonlinear iterative partial least squares (NIPALS) regression algorithm or a neural network, and calibration is performed, and the preprocessed signal intensity data are compared with viral titer data obtained under the same sample conditions using non-Raman spectroscopy such as qPCR, RT-qPCR, ELISA, or by visual determination by transmission electron microscopy, (e) Use model parameters obtained from each of the preprocessed datasets. The step includes using to estimate the response value, (f) Includes the steps of variable selection, optionally performing a Variable Importance Index (VIP) to identify Raman spectral variables, (g) Optionally, perform one or more further rounds of modeling on any dataset by reapplying steps (d) to (f), including the step of removing non-essential variables. This method provides a way to determine the ratio of viral nucleic acids to viruses containing one or more viral structural molecules in a sample, using model parameters obtained for identified Raman spectral variables derived from a multivariate data processing model. [Brief explanation of the drawing]

[0044] [Figure 1] Figure 1 is a Yablonski diagram showing quantum energy transitions for infrared absorption / emission. This diagram shows Rayleigh (elastic scattering) and Raman (inelastic scattering) in both Stokes transitions and anti-Stokes transitions. [Figure 2] Figure 2 shows an exemplary pre-treated Raman spectrum from a run of the VV Raman Project lentivirus bioreactor. The inset is a magnified view of the 1000 cm⁻¹ region. After a CCD readout time of approximately 15 minutes, spectra were acquired in 75 cumulative 10-second intervals, with a total integration time of approximately 12 minutes 30 seconds. [Figure 3] Figure 3 is a bar graph showing representative qPCR lentiviral titer results for bioreactor transfection in the Viral Vector Raman Project, where cp = copy number. [Figure 4] Figure 4 is a graph showing how the mean squared prediction error (MSEPCV) after 10-fold cross-validation and 20 Monte Carlo iterations changes as a function of the number of PLS ​​latent variables / components (before spectral variable selection). [Figure 5A-B] Figure 5A is a plot of the Variable Importance Index (VIP) calculated from the first 10 components of the PLS-R model. The circles represent spectral variables with a VIP > 1.5, and this information was used to determine the range shown in Figure 5B. [Figure 5C]Figure 5C is a table of spectral variables with VIP > 1.5, ranging in order of importance. [Figure 6] Figure 6 is a plot showing the change in mean squared prediction error after cross-validation as a function of latent variables or the number of components. It was obtained after conservative spectral variable reduction using VIP > 1.5. [Figure 7] Figure 7 is a plot showing the PLS-R prediction (10LV) of Raman copy number / mL from run 1 bioreactor 4 of lentivirus after reduction of spectral variables (VIP > 1.5) along with offline qPCR data. T means "transfected" and NT means "untransfected". [Figure 8] Figure 8 is a plot showing the PLS-R prediction (10LV) of Raman copy numbers from run 2 bioreactors 1-4 of lentiviruses after reduction of spectral variables (VIP > 1.5) along with offline qPCR data. [Figure 9] Figure 9 is a plot showing the PLS-R prediction (10LV) of Raman copy numbers from run 3 bioreactors 1-4 of lentiviruses after reduction of spectral variables, along with offline qPCR data. [Figure 10] Figure 10 shows a comparison of RT-qPCR results and p24 ELISA results. The p24 ELISA assay was used only to obtain lentiviral titers in some offline samples. [Figure 11]Figure 11 shows an example of applying a real-time Raman derivation model of viral titer to identify the initiation, production stages, and termination of the viral production process. From the model (solid black line), a set of three indicators is also calculated in real time. Along with the model estimate of viral titer, the shape of the indicator curves (as explained by the text overlaid on the graph) indicates the various stages of viral production. Physical titers, retrospectively calculated using either RT-qPCR or ELISA, or both, are overlaid on the graph (squares and dashed lines) to show the overall agreement between real-time data and retrospective offline data. While including additional peaks from Raman spectra could improve the quality of the Raman model, this plot was generated using the minimum number of regions necessary to identify the main stages of viral production. [Figure 12] Figure 12 shows a schematic diagram of the equation for quantifying viral titer. This equation is applied to model parameters (in this case, regression coefficients) obtained from a multivariate regression algorithm applied to normalized Raman signal intensity data. [Figure 13] Figure 13 shows a plot of R²(1 - sum of squared residuals / total sum of squares) as a function of the number of wavenumber ranges to indicate the minimum number of wavenumber ranges required to provide an estimate of the lentiviral titer. [Figure 14] Figure 14 shows an exemplary pre-treated Raman spectrum from a run of the AAV Raman Project bioreactor. The inset is a magnified view of the region 1000–1200 cm⁻¹. After a CCD readout time of approximately 15 minutes, the spectrum was acquired in 75 cumulative 10-second intervals, with a total integration time of approximately 12 minutes 30 seconds. [Figure 15] Figure 15 is a bar graph showing representative qPCR AAV virus titer results from bioreactor transfection in the AAV Raman Project. [Figure 16] Figure 16 is a graph showing how the mean squared prediction error (MSEPCV) after 10-fold cross-validation and 20 Monte Carlo iterations changes as a function of the number of PLS ​​latent variables / components (before spectral variable selection). [Figure 17A-B] Figure 17A is a plot of the Variable Importance Index (VIP) calculated from the first 15 components of the PLS-R model. The circles represent spectral variables with a VIP >= 1.0, and this information was used to determine the range shown in Figure 17B. [Figure 17C] Figure 17C is a table of spectral variables with VIP >= 1.0, ranging in order of importance. [Figure 18] Figure 18 is a plot showing the change in mean squared prediction error after cross-validation as a function of latent variables or number of components. The results were obtained after reduction of spectral variables using VIP >= 1.0. [Figure 19] Figure 19 is a plot showing the PLS-R prediction (9LV) of Raman copy numbers from run 4 bioreactors 1-4 after reduction of spectral variables, along with the offline qPCR data. [Figure 20] Figure 20 is a plot of R²(1 - sum of squared residuals / total sum of squares) as a function of the number of wavenumber ranges to show the minimum number of wavenumber ranges required to provide an estimate of the AAV virus titer. [Figure 21] Figure 21 shows an exemplary pre-treated Raman spectrum from a run of the AAV Raman Project bioreactor. The inset is a magnified view of the region 1000–1200 cm⁻¹. After approximately 15 minutes of CCD readout, spectra were acquired in 75 cumulative 10-second intervals, with a total integration time of approximately 12 minutes 30 seconds. [Figure 22] Figure 22 is a bar graph showing representative qPCR AAV virus titer results from bioreactor transfection in the AAV Raman Project. [Figure 23] Figure 23 is a bar graph showing representative ELISA AAV virus titer results for bioreactor transfection in the AAV Raman Project. [Figure 24] Figure 24 is a graph showing how the mean squared prediction error (MSEPCV) of the RT-qPCR copy number per ml changes as a function of the number of PLS ​​latent variables / components (before spectral variable selection) after 10-fold cross-validation and 20 Monte Carlo repeats. [Figure 25] Figure 25 is a graph showing how the mean squared prediction error (MSEPCV) of the RT-qPCR copy number per ml changes as a function of the number of PLS ​​latent variables / components (before spectral variable selection) after 10-fold cross-validation and 20 Monte Carlo repeats. [Figure 26A] Figure 26A is a plot of the Variable Importance Index (VIP) of the PLS-R model for copy number per mL (RT-qPCR-based), calculated from the PLS-R models of the first 15 components. [Figure 26B] The circles represent spectral variables where VIP >= 1.0, and this information was used to determine the range shown in Figure 26B. [Figure 26C] Figure 26C is a table of spectral variables with VIP >= 1.0, sorted by importance. [Figure 27A] Figure 27A is a plot of the Variable Importance Index (VIP) for the PLS-R model of total particles per mL (ELISA-based), calculated from the PLS-R models of the first 14 components. [Figure 27B] The circles represent spectral variables where VIP >= 1.0, and this information was used to determine the range shown in Figure 27B. [Figure 27C] Figure 27C is a table of spectral variables with VIP >= 1.0, ranging in order of importance. [Figure 28] Figure 28A is a plot of the Variable Importance Index (VIP) for the PLS-R model of total particles per mL (ELISA-based), calculated from the PLS-R model of the first 14 components. The circles represent spectral variables with VIP >= 1.0, and this information was used to determine the ranges shown in Figure 27B. Figure 27C is a table of spectral variables with VIP >= 1.0, ranging in order of importance. [Figure 29] Figure 29 is a plot showing the change in mean squared prediction error of the total viral particle count per ml of ELISA after cross-validation as a function of latent variables or component number. Obtained after reduction of spectral variables using VIP >= 1.0. [Figure 30]Figure 30 is a plot showing the Raman copy number PLS-R prediction (10LV) from bioreactors 1-4 after reduction of spectral variables, along with the offline qPCR data. [Figure 31] Figure 31 is a plot showing the Raman copy number PLS-R prediction (10LV) from bioreactors 5-8 after reduction of spectral variables, along with the offline RT-qPCR data. [Figure 32] Figure 32 is a plot showing the PLS-R prediction (10LV) of the total Raman particle number from bioreactors 1-4 after reduction of spectral variables, along with the offline ELISA data. [Figure 33] Figure 33 is a plot showing the PLS-R prediction (10LV) of the total Raman particle number from bioreactors 5-8 after reduction of spectral variables, along with the offline ELISA data. [Figure 34] Figure 34 is a plot showing the calculated Empty-Full Ratio (%) from Raman PLS-R model predictions of genome copy number (RT-qPCR) and total particle number (ELISA). It is shown for bioreactors 2-4, starting 24 hours post-transfection. [Figure 35] Figure 35 is a plot showing the calculated Empty-Full Ratio (%) from Raman PLS-R model predictions of genome copy number (RT-qPCR) and total particle number (ELISA). It is shown for bioreactors 5–7, starting 24 hours post-transfection. [Modes for carrying out the invention]

[0045] Detailed description of the invention It should be understood that different applications of the disclosed methods may be adapted to specific needs in the art. It should also be understood that the terminology used herein is intended solely to describe specific embodiments of the invention and is not intended to limit them.

[0046] Furthermore, as used herein and in the appended claims, the singular forms “a,” “an,” and “the” include plural references unless the content explicitly indicates otherwise.

[0047] All publications, patents, and patent applications cited herein, whether above or below, are incorporated herein by reference in their entirety.

[0048] Virus analysis using Raman spectroscopy The present invention encompasses the use of Raman spectroscopy for monitoring and evaluating viral titers and / or viral component abundances. The present invention encompasses the use of Raman spectroscopy for monitoring and evaluating viral nucleic acid abundances and viral structural molecule abundances. The present invention encompasses the use of Raman spectroscopy for monitoring and evaluating the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in a sample.

[0049] As defined herein, “viral titer” refers to the amount of virus present in a given volume. Any type of viral titer can be evaluated in this invention, for example, physical viral titer, functional viral titer (also called infectious viral titer), or transduction viral titer. In a particular embodiment, physical viral titer can be evaluated. Physical viral titer is a measure of the concentration of viral particles in a sample (e.g., a viral culture medium) and is usually based on the presence of viral proteins or viral nucleic acids, such as p24. Physical titer can be expressed as the number of viral particles per mL (VP / mL), the number of viral genomes per mL (vg / mL), the number of viral copies per mL, or the number of RNA copies per mL. Prior art assays for measuring physical titer include p24 ELISA (e.g., Lenti-X p24 Rapid Titer Kit (Takara) or Lentivirus-Associated p24 ELISA Kit (Cell Biolabs, Inc)), qPCR, or ddPCR (e.g., AAV Real-Time PCR Titration Kit (Takara) or Adeno X qPCR Titration Kit (Takara)). Measuring physical titer does not always distinguish between hollow or defective viral particles and particles that can infect cells. Therefore, physical viral titer can be distinguished from functional titer or infectivity titer, which determines how many of the produced particles can infect cells, and transdextrin titer, which determines how many of the functional viral particles contain the gene of interest (for example, the transdextrin titer may be relevant to the production of a viral vector). Thus, determining physical titer is not equivalent to determining functional titer unless all particles in the sample are functional. In fact, functional titer is often 1 / 100 to 1 / 1000 of physical titer.

[0050] Alternatively, as described above, functional or infectivity titer can be measured or evaluated in this invention, and functional or infectivity titer is a measure of the amount of viral particles present in a particular volume that can infect target cells. Functional titer can be expressed as plaque-forming units per mL (pfu / mL) or infectivity units per mL (ifu / mL). Offline assays that can be used to measure functional or infectivity titer include plaque assays, focal-forming assays, endpoint dilution assays, or flow cytometry. As described above, transduction titer is a measure of the amount of viral particles present in a particular volume that can infect target cells and contain the gene of interest. Transduction titer can be expressed as transduction units / mL and can be evaluated using assays used to evaluate functional titer above, along with any known assay that can determine the presence of the gene of interest, e.g., PCR. Those skilled in the art will understand that functional or transduction titer can be determined by reducing any value obtained for physical titer. As described above, physical titer and functional titer or transduction titer and The difference in magnification between these is well understood in the art. Therefore, in one aspect of the present invention, the functional or transducible titer can be determined indirectly by the method of the present invention (for example, by reducing the value obtained for the physical titer). Accordingly, the method of the present invention may include an additional step of determining the functional or transducible titer by reducing the determination of the physical titer.

[0051] The method of the present invention can be used to monitor and evaluate the amount of viral nucleic acid and viral structural molecules present in a sample (e.g., a viral culture medium). The method can also be used to determine the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in a sample (e.g., a viral culture medium). This ratio can be used to determine the proportion of viral particles in a sample that have both nucleic acid and structural components. This ratio can be used as an estimate of functional titer. Therefore, the method of the present invention can be used to estimate functional titer in real time, whereas previously known methods for estimating functional titer are retrospective and offline.

[0052] It will be understood by those skilled in the art that the method of the present invention can determine the viral titer and / or viral component abundance of any virus of any serotype, for example, retroviruses such as lentiviruses (e.g., HIV-1 and HIV-2) and gamma-retroviruses; adenoviruses and adeno-associated viruses (e.g., AAV1-11, particularly AAV1, AAV2, AAV5, and AAV8, and self-complementary AAVs). Therefore, the method of the present invention can be applied to mammalian viruses as further described and defined herein. The method of the present invention can be further applied to non-mammalian viruses, including plant viruses such as tobacco mosaic virus, algal viruses, yeast viruses, and insect viruses including baculoviruses, as further described and defined herein. Typically, the viral titer and / or viral component abundance of a single virus type can be evaluated by the present invention, but the method of the present invention can be used to evaluate the titer and / or viral component abundance of mixed virus samples, for example, samples containing two or more virus types.

[0053] In this regard, the Raman spectra generated according to the present invention can either directly detect the virus or viral component, indirectly detect the virus or viral component, or detect the virus or viral component both directly and indirectly. Therefore, the wavenumber peaks generated and shown in the Raman spectrum may indicate one of several different compounds or molecules associated with the virus or viral component. Those skilled in the art will understand that for the performance of the present invention, it is not necessary to precisely identify which compound / molecule each identified wavenumber peak corresponds to. Instead, the spectrum functions as a fingerprint under specific conditions (e.g., culture conditions, virus, producer cell line, and type of viral titer to be measured), and the viral titer and / or viral component abundance can be determined by analyzing the signal intensity at each of the wavenumber ranges disclosed herein using a specific multivariate model (typically generated under identical or similar conditions). Multivariate models are described in detail herein. Therefore, peaks obtained for a specific Raman spectrum at a specific wavenumber may correspond to viral molecules / compounds (e.g., capsid proteins) or to non-viral molecules / compounds (e.g., metabolites in cultures produced by virus-producing cells). Thus, the Raman spectroscopy used in the present invention can detect not only compounds directly related to viral titer and / or viral component abundance, but also compounds indirectly related to viral titer and / or viral component abundance.

[0054] The signal strength may differ in each wavenumber range, and under different conditions (e.g., different viruses) While Raman spectra obtained under different conditions (such as cell production, use of different producer cell lines, or use of different culture media or bioreactors) may correspond differently to viral titers and / or viral component abundances, the wavenumber range that can be evaluated (e.g., those determined to be related to viral titers and / or viral component abundances) is expected to remain the same. Furthermore, for any given set of conditions, the user will understand that a multivariate model can be constructed based on the principles described herein and known in the art, and this model can be applied when analyzing different signal intensity data from the same wavenumber range described herein and subsequently generated using the same set of conditions. This means that the user can calculate viral titers and / or viral component abundances via Raman spectroscopy using the wavenumber range described herein, and the data is generated in systems where different conditions are applied.

[0055] The wavenumber ranges used herein for lentiviral assessment have identified ranges that are important in assessing viral titer and / or viral component abundance. Because lentiviruses are known to be particularly complex in terms of their chemical composition, the chemical components of simpler viruses that possess some of the components present in lentiviruses may generate relevant signals in a subset of the identified wavenumbers, e.g., at least five of the identified wavenumber ranges (if one or more of the wavenumber ranges directly detect the virus). Therefore, the subset of wavenumber ranges provided herein can be used to assess the titer of non-lentiviral viruses. Furthermore, if one or more wavenumber ranges identified as correlated with viral titer and / or viral component abundance indirectly detect the virus, such wavenumber ranges may be useful in assessing viral titer and / or viral component abundance in any system, since viral transfection is likely to induce similar metabolic changes in cultures.

[0056] In a particular embodiment of the present invention, lentiviral titer and / or lentiviral component abundance are monitored and evaluated. In one embodiment, the virus does not have to be HIV-1 or HIV-1 virus-like particles.

[0057] A “virus” is typically a small infectious pathogen (usually smaller than a bacterium) that can only replicate within the living cells of another organism. Viruses may have an RNA or DNA-based genome. As used herein, “virus” refers to any virus, modified virus, viral particle, virus-derived particle, or viral vector. Thus, while the viral titer and / or viral component abundance of any wild-type virus can be evaluated according to the present invention, it will be understood that the usefulness of the present invention may extend particularly to the evaluation of the viral titer and / or viral component abundance of mutant or modified viruses (i.e., those that involve one or more nucleic acid substitutions, insertions, deletions, or transpositions, or lack a significant portion of the genetic material encoding viral proteins, compared to wild-type or naturally occurring viruses) or viral vectors. Mutant or modified viruses, or viral particles, are often used to manufacture vaccines, and the methods of the present invention are considered particularly effective for monitoring and evaluating the production efficiency of functional viruses or viral particles for use in vaccines.

[0058] Viral vectors can be based on wild-type viruses, but they are generally modified compared to wild-type viruses and are commonly used to introduce genetic material into target cells (e.g., genes for therapeutic purposes). Therefore, viral vectors have specific utility for applications such as gene therapy, cell therapy, or other molecular applications, and their production is crucial to the gene therapy and cell therapy industries. For example, modifications may be made to improve the safety of viral vectors for gene and / or cell therapy, or, for example, to improve the size of the genes that can be carried by the vector. It will be understood that modifications that can be made to create a viral vector may include deletions of parts of the viral genome that are important for replication, resulting in a viral vector that can infect cells, but this requires the presence of a helper virus to provide the missing protein necessary for the production of new virions. Other modifications may include modifications to reduce genome rearrangement, for example, to reduce the toxicity of the viral vector to its target cells and / or to improve the stability of the virus.

[0059] Viral vectors can typically be produced by transducing a packaging cell line, such as HEK293 cells, with one or more plasmids that encode viral proteins and carry the required genetic material. For example, for lentiviral production, HEK293 cells can be transduced with one or more plasmids, e.g., three or four plasmids that encode virion proteins such as capsids or reverse transcriptase and carry the genetic material to be delivered by the vector. This is transcribed to produce a single-stranded RNA viral genome, marked by the presence of a psi sequence, which ensures that the genome is subsequently packaged into virions. Thus, lentiviral vectors can be produced in particular by the transformation and expression of three (in the case of a second-generation system) or four (in the case of a third-generation system) plasmids in the producer cell line. Plasmids for producing viral vectors are commercially available, e.g., Lenti-Pac and AAVPrime (GeneCopoeia).

[0060] In particular, the titer and / or viral component abundance of viral vectors produced by packaging cell lines can be monitored or evaluated by the method of the present invention. As mentioned above, in the fields of gene therapy and cell therapy, the ability to measure the generated titer and / or viral component abundance with high sensitivity is especially important, as it allows for accurate monitoring and control of production processes, such as production from producer cell lines, and real-time calculation of the proportion of functional vectors. The virus does not need to be fully functional or wild-type to be monitored or evaluated by the method of the present invention.

[0061] In this specification, "viral component" is considered to be any part of a virus, viral particle, or viral vector.

[0062] A viral particle, or "virion," has traditionally been understood to consist of (i) viral genetic material, i.e., a molecule of DNA or RNA that codes for the structure of the protein on which the virus acts; (ii) an internal protein membrane called a capsid, formed from a capsomere, that surrounds and protects the viral genetic material; and, in some cases, (iii) a lipid outer envelope that may contain envelope proteins.

[0063] The viral components include viral nucleic acids and viral structural components (or viral structural molecules). Viral nucleic acids are considered, in this specification, to include viral RNA, viral DNA, viral DNA genome, and viral RNA genome. The viral nucleic acids are packaged within the virion.

[0064] As used herein, viral structural components or viral structural molecules should be understood as any molecule that contributes to the structure of a virus. As used herein, viral structural components or viral structural molecules may exclude viral genetic material, i.e., DNA or RNA molecules that encode the structure of the protein on which the virus acts.

[0065] Viral structural components or viral structural molecules are, in this specification, viral proteins (e.g., nucleoproteins; capsid proteins; protomers; capsid subunits; capsids) It is thought to include a combination of domeromers; capsidomeres; capsomeres; hexones; pentons; viral coat proteins (VCPs); viral outer surface glycoproteins; viral transmembrane proteins; proteins essential for the function of viruses, viral particles, or viral vectors), viral glycans, glycosylated viral molecules (e.g., glycosylated viral proteins), and / or viral lipids including viral phospholipids, or combinations thereof.

[0066] As described above, the method of the present invention can "monitor or evaluate" viral titer and / or viral component abundance. Accordingly, the method of the present invention can determine viral titer and / or viral component abundance, for example, the level, amount, or concentration of viral nucleic acid and viral structural molecules present in a sample. Accordingly, in particular, the method can determine whether the level, amount, or concentration of viral nucleic acid and viral structural molecules increases or plateaus over time (for example, by assaying samples at different time points) or changes (e.g., increases, decreases, or remains the same) compared to different samples (e.g., assayed at the same or equivalent time points). Thus, the method of the present invention can be used, for example, to evaluate the efficiency of a method of producing a virus, where the detection or determination of the ratio of viral nucleic acid to a virus containing one or more viral structural molecules may indicate an efficient or suboptimal production method, or, for example, to determine the importance of a particular factor in a method of producing a virus by comparing it with viral titer and / or viral component abundance measured between other modified production methods (for the same or different viruses). The physical titer expected to be detected by the method described herein is 1 × 10⁻⁶ 10 ~1 × 1011 The range is particles / mL. The infectivity titer expected to be detected by the method described herein is 1 × 10⁻⁶. 8 ~1 × 10 9 The range is particles / mL. Therefore, modifications of factors that result in differences in the measured viral titer and / or viral component abundance (e.g., modifications of factors that result in a difference of at least 5, 10, 20, 30, 40, or 50% in the measured viral titer) can be determined to be important to the production method. Such factors may include incubation temperature, culture medium used, % glucose or amino acids used in the medium, presence or absence or amount of stirring used, or culture flask or volume used.

[0067] Therefore, using the method of the present invention, it is possible to determine the optimal conditions for virus production, including the evaluation of different systems available for culturing producer cells that may produce the virus, such as shaker flasks, Quantum systems (Terumo), and Ambr systems (e.g., Ambr 15 or 250 (TAP Biosystems)).

[0068] The method of the present invention can be further used to evaluate any process downstream of a virus production process, for example, to determine whether such a process affects the viral titer and / or the abundance of viral components. In particular, the method of the present invention can be used to evaluate a purification method that may be used, for example, to determine whether such a purification method has any effect on the titer and / or the abundance of viral components, for example, by comparing the ratio of viral nucleic acid to virus containing one or more viral structural molecules that was present in the sample before purification with the ratio of viral nucleic acid to virus containing one or more viral structural molecules that was present after purification, to determine whether the ratio increased, decreased, or remained the same. The method of the present invention can be further used to evaluate the large-scale production of viruses, for example, viral particles for use in vaccines, or viral vectors that may be particularly important for the production of viral vectors for gene therapy.

[0069] As used herein, an increase in viral titer and / or viral component abundance may be an increase exceeding 5, 10, 20, 30, 40, 50, 60, 70, 80, or 90% of the viral titer and / or viral component abundance being compared, and as used herein, a decrease in viral titer and / or viral component abundance may be a decrease exceeding 5, 10, 20, 30, 40, 50, 60, 70, 80, or 90% of the viral titer and / or viral component abundance being compared. An equivalent viral titer and / or viral component abundance may be within 5% of the viral titer being compared.

[0070] In this regard, it will be understood that for several purposes, it may be desirable to evaluate the viral titer and / or viral component abundance before and after and / or during the method in order to determine whether a change or fluctuation has occurred in the viral titer and / or viral component abundance. The method of the present invention may also further include a step of comparing the viral titer and / or viral component abundance with, for example, the viral titer and / or viral component abundance in different samples (at equivalent or different time points) or in the same sample at different time points.

[0071] In a further embodiment of the present invention, the method may be used to determine the extent of viral infection in a subject, for example, to determine whether the infection is being successfully treated or mitigated. In such a method, it may be desirable to compare the viral titer and / or viral component abundance in the sample with that of the same type of sample from a subject at different time points in order to determine whether the ratio of viral nucleic acid to virus containing one or more viral structural molecules is increasing, decreasing, or remaining constant over time. Alternatively or additionally, it may be desirable to compare the viral titer and / or viral component abundance in a sample from a particular individual with previously obtained measurements of viral titer and / or viral component abundance that may indicate, for example, the stage of infection and / or prognosis for a particular condition.

[0072] Alternatively, the method of the present invention may not determine the actual amount, level or concentration of viral component abundance in a sample, but may determine whether the amount, level or concentration is above or below an acceptable threshold. For example, in a certain production method, the threshold can be used to determine whether there is an acceptable level of functional viral particles in the sample. As described above, the method of the present invention can determine whether the ratio of viral nucleic acid to virus comprising one or more viral structural molecules is increased, decreased, or comparable relative to the ratio of a previously assayed sample, and it will therefore be appreciated that for certain specific applications, it may not be necessary to determine the actual viral titer and / or viral component abundance (e.g., the amount or concentration of virus present).

[0073] The present invention encompasses the use of Raman spectroscopy to monitor and / or evaluate viral titer and / or viral component abundance in a sample, and thus can identify any one of the start, production phase, and end of a virus production process. It will be appreciated that different amounts or concentrations of virus and / or metabolites are present in a sample at different stages of production. For example, at the start of production, the viral titer, particularly physical viral titer, is 0 to 10 5 and during active virus production, the viral titer, particularly physical viral titer, is 10 5 to 10 9 and at the end of virus production, the viral titer, particularly physical viral titer, is 10 9 to 10 12 However, generally, monitoring of viral titer and / or viral component abundance over time shows that, for example, increased or peak amounts may be associated with the production phase of the process, early small amounts may be associated with the start of the process, and subsequent stagnation of the amount may be associated with the end of the process for a specific package It is possible to identify different stages of production of a particular virus in a cell line. As previously described, this information can be used in specific processes to ensure that the culture is not maintained after production has stalled, decreased (e.g., decreased by at least 50% compared to the peak point of production), or terminated.

[0074] The method of the present invention can also be used to support adaptive production and further increase the production of viral titer and / or viral component abundance in a system. In this regard, the ratio of viral nucleic acid to a virus containing one or more viral structural molecules obtained using different systems under different conditions can be compared to determine the optimal conditions for viral production.

[0075] As used herein, the term “sample” refers to any sample containing a virus (e.g., any sample containing a viral vector). The “sample” is preferably a viral culture medium, i.e., a liquid in which the virus is incubated. Thus, the viral culture medium can be directly irradiated to obtain Raman spectroscopic data for use in the present method, as further described herein. In a preferred embodiment of the present invention, the sample is from an industrial virus production process or a viral vector manufacturing process. In particular, in the present invention, the viral titer in the sample can be measured in real time in situ by Raman spectroscopy, or ex situ on the sample.

[0076] "In situ" means that the measurement for obtaining the intensity of Raman scattering light in a culture capable of producing viral particles is taken from the primary culture environment in which the viral particles are generated, rather than from a sample extracted from the primary culture environment. Therefore, performing the measurement "in situ" eliminates the requirement of a liquid handling step. Thus, removal of the sample from that environment may not be necessary for certain applications of the present invention, and in-situ measurement of the sample may be preferred. In-situ measurement of a sample can enable periodic assessment of viral titer and / or viral component abundance in a sample without requiring an actual sampling step in which a portion of the sample (e.g., viral culture medium) is removed from the primary culture environment (e.g., viral growth incubator). Assessment of viral titer and / or viral component abundance in this respect can be measured accurately and sensitively in real time without requiring additional steps that could introduce cost and errors. As will be described in more detail below, Raman spectroscopy provides probes that can be placed either inside or outside the sample, for example, enabling in-situ measurements as needed. In-situ measurement is particularly well suited to "in-line" process analysis techniques.

[0077] Alternatively, the method of the present invention can be performed on a sample in excitually. "Excitually" means that the measurement for obtaining the intensity of Raman scattering light in a culture capable of producing viral particles is obtained from an aliquot of a sample (e.g., viral culture medium) extracted from the primary culture environment (e.g., a viral growth incubator) where the viral particles are produced, and analyzed directly. Such excitually measurements are suitable for "atline" or retrospective process analysis techniques. Whether the measurement is performed in situ or excitually, the measurement can be performed directly on the sample (e.g., directly on the viral culture medium) without requiring further processing of the sample.

[0078] The origin of the sample used in the method of the present invention may be a cell culture in which the virus is produced. Therefore, the sample is one of the culture media (e.g., DMEM, MEM, or SFII, optionally containing serum, L-glutamine, and / or other components). This can include, for example, packaging cells (e.g., HEK293 cells) if collected during the virus production process, or it can be a sample of the virus for medical use, for example, requiring quality testing before marketing, selling, or use. The sample can further be a sample from a subject suspected of being infected with the virus (e.g., a human or mammalian subject), such as a blood, saliva, sputum, plasma, serum, cerebrospinal fluid, urine, or fecal sample. Other sources of the sample include those from open water or public water.

[0079] Raman spectroscopy Raman spectroscopy measures the change in wavenumber of monochromatic light scattered by a sample to provide information about the sample's chemical composition, physical state, and environment. This is possible because incident photons interact with vibrational modes present in the molecules that make up the sample. These modes have specific vibrational frequencies and scattering intensities under a given set of physical conditions, thereby allowing for the quantification of the quantity of a given analyte. Unlike infrared absorption spectroscopy, which measures the absorption of light of different energies from a broadband light source, Raman spectroscopy measures the energy difference between the monochromatic incident and scattered light (Figure 1), which is known as the Raman shift.

[0080] Typically, Stokes scattered light is monitored so that the measured signal is stronger at ambient temperature. Figure 2 shows several exemplary spectra, where different peaks represent the presence of different vibrational modes, and some bands are overlapping regions of several underlying peaks. For simple mixtures, distinct bands specific to the n-1 analyte can be identified, and the change in intensity at constant temperature and pressure can be measured to fully quantify the composition after proper calibration. However, for complex systems such as biological media, this simple approach cannot be strictly applied, as there are too many overlapping bands, making direct assignment impossible. Instead, in addition to modeling the original data matrix, it is necessary to use advanced chemometric models that obtain a linear combination of variables (wavelength / wavenumber) that maximize the covariance with the concentration of the analyte in question. The composition of a new sample can then be predicted.

[0081] Raman spectroscopy provides a "molecular fingerprint," enabling qualitative and quantitative analysis of samples, such as biological samples. Raman spectra are generally sensitive to changes in physical conditions, such as temperature and pH. Often, Raman spectra obtained from biological samples may contain background fluorescence signals, as such samples frequently contain natural fluorophores. In conventional Raman spectroscopy, this background should be limited by the selection of an optimal laser wavenumber and the removal of the remaining fluorescence using one of several conventionally available algorithms.

[0082] Raman spectroscopy is a well-known technique in the art. In this invention, real-time Raman spectroscopy can be used in situ as described above, enabling continuous measurement of viral titers and / or viral component abundances.

[0083] As used herein, “Raman spectroscopy” can refer to all types of Raman spectroscopy that do not require binding between the substrate and the target molecule of interest (e.g., the molecule detected by Raman), such as an immunological interaction. Binding between the substrate and the target molecule may occur directly or indirectly using any type of binding molecule, such as streptavidin / biotin, or antibody fragments. “Immune interaction” includes the use of an antibody or antibody fragment (e.g., scFvs) that can be attached to the substrate to specifically bind to the target molecule in the sample. In particular, as used herein, “Raman spectroscopy” is Surface-enhanced Raman spectroscopy (SERS), for example, SERS that require immunoreactions with a substrate (which may include, for example, metal (e.g., Au) nanodots), can be excluded. SERS requires the analyte to be detected to be immobilized on a surface. In one embodiment of the method of the present invention, Raman spectroscopy is not performed to detect analytes in a sample immobilized on a surface. In particular, in one embodiment of the present invention, Raman spectroscopy is not performed to detect viral particles in or from a sample immobilized on a surface.

[0084] Unlike Raman spectroscopy according to a preferred embodiment of the present invention, SERS requires a specific experimental design for immobilizing or binding the analyte of interest to a surface, which results in enhanced signal intensity using the SERS methodology. However, such immobilization of the analyte requires sample processing, which can lead to contamination of or interference with the conditions inside the bioreactor, including the circumstances under which the sample is taken from such a system. Generally, SERS is more suitable for methods involving "simple" samples containing the analyte of interest, where the sample contains few contaminants, than for complex mixtures of components such as those found in bioreactors or biological samples for processing viruses according to the methods defined herein. Thus, conventional SERS is not ideally designed for direct monitoring or in-line or in-situ monitoring of complex samples, including samples containing viral particles for evaluation of viral titer, but more typically, it can be applied to the analysis of processed samples where the analyte to be detected has been purified and subsequently immobilized on a surface.

[0085] In contrast, preferred embodiments of the method of the present invention using Raman spectroscopy as defined herein can detect analytes, particularly viral particles, in a sample in-line / in situ without surface adhesion of analytes present in the sample, particularly without surface adhesion or immobilization of viral particles.

[0086] Raman spectroscopy as defined in this invention includes, in particular, conventional types of Raman spectroscopy, as well as other types of Raman spectroscopy such as stimulated Raman spectroscopy (SRS), pico-Raman, spatial offset Raman (SORS), inverse SORS, see-through Raman spectroscopy, coherent anti-Stokes Raman spectroscopy (CARS), coherent Stokes Raman spectroscopy (CSRS), resonant Raman spectroscopy (RR spectroscopy), and total internal reflection Raman spectroscopy (TIR) ​​Raman. Instruments for Raman spectroscopy are available from various suppliers, e.g., Renishaw, WITec, Horiba, and ThermoFisher Scientific. See also http: / / www.optiqgain.com / , https: / / www.timegate.com / , and https: / / www.newport.com / .

[0087] Data processing and multivariate modeling As used herein, “multivariate” data refers to data in which multiple variables are measured for each sample, and “multivariate model” is a model constructed using such multivariate data. Raman spectra (e.g., generated from cell cultures over a period of time) contain multivariate data and can record intensities at multiple wavenumbers for each sample or time point measured.

[0088] In this invention, Raman spectra and multivariate data constituting the spectra obtained from in-situ monitoring of virus production in cultures were analyzed to identify a set of spectral variables that are most important for enabling model prediction to achieve real-time measurement of viral titer and / or viral component abundance. In particular, the model prediction evaluates the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in the culture. To achieve this, the abundance of viral nucleic acid and viral structural molecules was measured. Plots of variable importance indices (VIPs) calculated from 10 or 12-component multivariate models were prepared to establish the importance of wavenumber variables, for example, with respect to the data shown in Table 3. Plots of variable importance indices (VIPs) calculated from 8 or 15-component multivariate models were also prepared to establish the importance of wavenumber variables, for example, with respect to the data shown in Table 1. Accordingly, the inventors used multivariate data from Raman spectral measurements, along with offline data related to viral titer and viral nucleic acid abundance or viral structural molecule abundance from prior art assays, to identify wavenumber ranges that can be evaluated when determining viral titer and / or viral component abundances, and further constructed multivariate models that can analyze signal intensity in identified wavenumber ranges from any Raman spectra achieved for virus-containing samples under specific conditions to determine viral titer and / or viral component abundances. Any type of viral titer and / or viral component abundance described herein can be determined using the method of the present invention, provided that the multivariate model used to predict viral component abundance is constructed using offline data on the relevant type of viral component abundance.

[0089] In complex biological samples, clearly assigning wavenumber ranges to specific analytes is often difficult or impossible. Low concentrations of analytes also present problems, as signals can be obscured by other high concentrations of compounds such as water or glucose in the cell culture medium. If the obscured signals are too strong, the associated noise is greater than the measured variation originating from the underlying target analyte, causing the signal intensity of the target component to fall below the detection limit. Through Raman spectral analysis, modeling of acquired data, and calibration of data by offline measurements of viral titer and viral nucleic acid abundance or viral structural molecule abundance, the inventors identified a range of wavenumber ranges that correlate with increases in viral titer and / or viral component abundance in the sample, particularly the ratio of viral nucleic acid to virus containing one or more viral structural molecules in the sample. These wavenumber ranges show a strong, consistent correlation with viral titer and / or viral component abundance over time, eliminating the need to assign ranges to specific analytes. Therefore, the problem of signal obscuration by high concentrations of compounds is not as significant in the method of the present invention.

[0090] Multivariate model parameters were obtained to determine the wavenumber range for use in the present invention. These parameters were then used to select key variables to be used in subsequent analyses to estimate response values ​​and to calculate viral titers, viral nucleic acid abundances, and viral structural molecule abundances. In this case, regressions were performed on Raman data obtained from virus-containing samples.

[0091] In this particular case, the inventors performed regressions on pre-processed Raman data. In this regard, it will be understood that raw Raman spectral data obtained from a spectrometer often require correction of several interference signals, such as background fluorescence, and it is often important to normalize the raw spectrum obtained for a sample to correct for the overall change in absolute intensity. Therefore, those skilled in the art will understand that it may be necessary to perform spectral pre-processing to address such issues.

[0092] Preprocessing of acquired Raman spectra can be performed using one of many algorithms available in the scientific literature. In particular, first-order differentiation, second-order differentiation (Savitzky et al., 1964), and standard normalized variable (SNV) normalization, as well as polynomial background fitting and elimination, can be used, for example. Barnes et al. (1989) describe the standard normalized variable method. Lieber and Mahadevan-Jan Sen (2003) describes an automated method for subtracting fluorescence from biological Raman spectra based on a modified least-squares polynomial curve fitting. Zhao et al. (2007) describe an improved automated algorithm for fluorescence removal based on modified multiple polynomial fittings. Koch et al. (2017) describe a “softener”-based baseline correction algorithm for Raman spectrum preprocessing. Hu et al. (2018) describe a baseline correction method based on piecewise polynomial fitting (PPF). Zhang et al. (2010) describe an adaptive, iteratively reweighted, penaltyly least-squares (airPLS) algorithm. See also Huang et al. (2010).

[0093] Once the steps for pre-processing the Raman spectra are performed, parameters for data modeling are obtained as needed. In this regard, regression of Raman data, particularly pre-processed Raman data, can be performed based on offline responses obtained using other techniques such as qPCR and p24 ELISA, plaque assays, which can determine viral titers, or CuBiAn and LC-MS, which can be used for the analysis of metabolic markers. Thus, regression may involve comparing pre-processed Raman spectra with offline data. A typical approach to multivariate regression is to use partial least squares regression (PLS-R).

[0094] Specifically, in standard orthogonal score PLS-R analysis, a linear relationship is obtained between the sequence of the Raman spectrum X and the response y (for example, y can be a vector where each element represents the titer or viral component abundance of each sample: y = α + Xβ + E In the formula, α and β are unknown parameters, and E is the error intensity or residual matrix. Depending on whether y contains a single response value or multiple response values ​​for each sample, different basic PLS-R algorithms can be used; the PLS1 algorithm can be used for a single response value, and the PLS2 algorithm can be used for multiple response values. When predicting viral titer and / or viral component abundance from Raman spectra, y is usually a univariate parameter for each sample, so PLS1 can usually be used.

[0095] In short, the PLS1 algorithm works as follows: First, the variables X and y have their mean centered (the mean of each variable can be subtracted from each element of the column of X, and the mean of y can be subtracted from each element of y). Some underlying factor A can be chosen to model X for a model, which is a factor that can be used in a linear combination. In the first step, we can find the weights by projecting X onto y, and these weights define the direction of the vector / factor space of X that has the largest covariance with y. Next, these weights are normalized to have unit length. Then, the X score can be calculated by projecting X onto these normalized weights. Next, the loadings of X can be calculated by projecting X onto the score. Similarly, the loadings of y are These scores can be calculated by projecting the transpose of y onto them. The contribution of the current component can be removed from both X and y by contracting X and y by subtracting the contribution of the given component. This can be done by multiplying each component's score by its loading vector and subtracting the resulting array or vector from the X array and y vector, respectively. The contracted X and y can then be reused in the same way for each subsequent component of the iterative procedure, that is, the weights, scores, and loadings are determined sequentially until all A components are used up and no further contraction is performed. A reference to this PLS method and the NIPALS algorithm can be found, for example, in Wold et al. (2001).

[0096] During each iteration, the calculated weight, score, and loading vectors and scalars are sequentially stored in their own arrays or vectors, that is, each iteration, the associated vectors Alternatively, scalars can be placed as new columns or rows within an array or element in an existing vector or vector, where the scalars may have been obtained from previous iterations. If regression coefficients are used as parameters for subsequent data modeling, the regression coefficient β can be obtained by multiplying the final weight matrix by the reciprocal of the projection of the transpose of the final X-block loading array. The optimal number of components can be determined by examining the prediction errors of a test set of preprocessed Raman spectra not used in the iterative model building procedure.

[0097] The prediction of viral titer and / or viral component abundance can be performed as described above, using model parameters such as the regression coefficient β and comparing them with offline assay data, e.g., the mean squared error (MSE) of a qPCR / P24 ELISA / plaque assay. The optimal number of underlying component A can be selected when the predicted MSE is minimized. For viral nucleic acid abundance, offline assays including qPCR or RT-qPCR can be used. For viral structural molecule abundance, offline assays such as ELISA can be used.

[0098] For example, after building the initial PLS-R model using the procedure described above, the most important variables can be selected using one of the many available variable selection methods, such as a Variable Importance Index (VIP) that identifies the variables most likely to be important in explaining the predictions of y and the variances of X. VIP can generate a VIP vector of the same length as the rows of X, meaning the VIP vector contains elements corresponding to each variable in X, where the numerical value of each element can be a measure of the importance of that variable. A common approach is to refine the initial model described above by rebuilding the model using only the most important variables, as determined by VIP or other variable selection methods. A threshold approach is often chosen to determine which variables are important. The VIP threshold can typically be set to 1, as it is the mean value of the VIP parameters, but those skilled in the art will understand that this is essentially arbitrary and other higher thresholds, such as 1.5, can be chosen.

[0099] As discussed below, the wavenumber range identified in the present invention as important for determining the viral titer and / or viral component abundance is based on setting the VIP parameter to at least 1.00 or higher. Therefore, in the present invention, wavenumber ranges that did not produce a peak intensity greater than 1.00 at this stage were excluded.

[0100] Those skilled in the art will understand that the PLS1 algorithm can be run again after the selection of the VIP parameters, in which case the variables of X that fall below the VIP threshold can be removed to generate a new multivariate model with shorter loading vectors and shorter regression coefficient β vectors.

[0101] The wavenumber range shown below is the result of Raman spectroscopy performed using a laser with a wavelength of 785 nm. The present invention includes the possibility of using lasers of different wavelengths other than 785 nm. The wavenumber ranges obtained using lasers of different wavelengths will be the same because the Raman shift (i.e., the difference in wavelength of inelastically scattered Raman light from a monochromatic laser beam used to induce Raman scattering) is largely wavelength-independent.

[0102] "Wavenumber" as used herein JPEG2026143417000001.jpg2117 This is the number of wavelengths per unit distance, and is in cm. -1 It is measured in units. Usually, JPEG2026143417000002.jpg1844 However, λ is the wavelength.

[0103] In a preferred embodiment of the present invention, peaks greater than 1.5 (Variable Importance Index (VIP)) indicate wavenumber ranges that are likely to be important for determining the presence of the virus and the viral titer and / or viral component abundance. Determining the signal intensity of specific wavenumber ranges is essential for predicting viral titer by the method of the present invention. The 1.5 VIP threshold is used in one embodiment of the present invention to determine which wavenumber ranges are important for predicting viral titer and / or viral component abundance. Accordingly, in a preferred embodiment, the present invention encompasses a method for evaluating or predicting viral titer and / or viral component abundance using Raman spectroscopy, comprising the step of determining the signal intensity at five or more of these wavenumber ranges from the Raman spectrum obtained from the sample. The measured preprocessed signals are used for prediction. Any subsequent spectra, i.e., spectra after model construction in which prediction is made, require preprocessing using the exact same method as the data used for training and building the PLS model.

[0104] Any method of the present invention includes the step of measuring the total intensity of Raman scattered light in each of a plurality of wavenumber ranges to obtain a set of wavenumber intensity data of a sample, wherein the plurality of wavenumber ranges are pre-selected and characteristic of the viral components in the sample.

[0105] In any of the methods of the present invention, the multiple wavenumber ranges in the measured Raman spectrum may include four or more wavenumber ranges 1 to 12, all of which have a VIP of 1.00 or greater, as listed in Table 1 below. The multiple wavenumber ranges in the measured Raman spectrum may include six or more wavenumber ranges 1 to 12, all of which have a VIP of 1.00 or greater, as listed in Table 1 below. The multiple wavenumber ranges in the measured Raman spectrum may include eight or more wavenumber ranges 1 to 12, all of which have a VIP of 1.00 or greater, as listed in Table 1 below. The multiple wavenumber ranges in the measured Raman spectrum may include ten or more wavenumber ranges 1 to 12, all of which have a VIP of 1.00 or greater, as listed in Table 1 below. The multiple wavenumber ranges in the measured Raman spectrum may include all twelve wavenumber ranges 1 to 12, all of which have a VIP of 1.00 or greater, as listed in Table 1 below.

[0106] In any of the methods of the present invention, the multiple wavenumber ranges in the measured Raman spectrum may include four or more wavenumber ranges from 13 to 22, where VIP is 1.25 or greater, as listed in Table 1 below. The multiple wavenumber ranges in the measured Raman spectrum may include six or more wavenumber ranges from 13 to 22, where VIP is 1.25 or greater, as listed in Table 1 below. The multiple wavenumber ranges in the measured Raman spectrum may include eight or more wavenumber ranges from 13 to 22, where VIP is 1.25 or greater, as listed in Table 1 below. The multiple wavenumber ranges in the measured Raman spectrum may include all ten wavenumber ranges from 13 to 22, where VIP is 1.25 or greater, as listed in Table 1 below.

[0107] In any of the methods of the present invention, the multiple wavenumber ranges in the measured Raman spectrum are wavenumber ranges 23-30, where VIP is 1.50 or greater, as listed in Table 1 below. It can include four or more of these. Multiple wavenumber ranges in the measured Raman spectrum can include six or more of the wavenumber ranges 23-30 with a VIP of 1.50 or greater, as listed in Table 1 below. Multiple wavenumber ranges in the measured Raman spectrum can include all eight of the wavenumber ranges 23-30 with a VIP of 1.50 or greater, as listed in Table 1 below.

[0108] [Table 1]

[0109] The viral titer and / or viral component abundance can be measured using multiple wavenumber ranges in a Raman spectrum, as described above in relation to Table 1. In one preferred embodiment of the present invention, the viral titer and / or viral component abundance measured using multiple wavenumber ranges in a Raman spectrum, as described above in relation to Table 1, is the adeno-associated virus (AAV) titer. In one particularly preferred embodiment of the present invention, the viral titer and / or viral component abundance measured using multiple wavenumber ranges in a Raman spectrum, as described above in relation to Table 1, is the adeno-associated virus serotype 8 (AAV8) titer. In any of the methods of the present invention, the multiple wavenumber ranges of the measured Raman spectrum may include four or more wavenumber ranges 1 to 20, as listed in Table 2, where VIP is 1.00 or greater, or the multiple wavenumber ranges of the measured Raman spectrum may include six or more wavenumber ranges 1 to 20, as listed in Table 2, where VIP is 1.00 or greater, or the multiple wavenumber ranges of the measured Raman spectrum may include four or more wavenumber ranges 1 to 20, as listed in Table 2, where VIP is 1.00 or greater. It can include 8 or more of the above, or the multiple wavenumber ranges of the measured Raman spectrum can include 10 or more of the wavenumber ranges 1 to 20 with VIP of 1.00 or greater, as listed in Table 2, or the multiple wavenumber ranges of the measured Raman spectrum can include 12 or more, 14 or more, 16 or more, or 18 or more of the wavenumber ranges 1 to 20 with VIP of 1.00 or greater, as listed in Table 2, or the multiple wavenumber ranges of the measured Raman spectrum can include 8 or more of the above, as listed in Table 2 This means that the VIP is 1.00 or higher, and it can include all 20 values ​​within the wavenumber range of 1 to 20.

[0110] In any of the methods of the present invention, the multiple wavenumber ranges of the measured Raman spectrum may include four or more of the wavenumber ranges 21 to 33 listed in Table 2, where VIP is 1.25 or greater; or the multiple wavenumber ranges of the measured Raman spectrum may include six or more of the wavenumber ranges 21 to 33 listed in Table 2, where VIP is 1.25 or greater; or the multiple wavenumber ranges of the measured Raman spectrum may include eight or more of the wavenumber ranges 21 to 33 listed in Table 2, where VIP is 1.25 or greater; or the multiple wavenumber ranges of the measured Raman spectrum may include ten or more, eleven or more, or twelve of the wavenumber ranges 21 to 33 listed in Table 2, where VIP is 1.25 or greater; or the multiple wavenumber ranges of the measured Raman spectrum may include all thirteen of the wavenumber ranges 21 to 33 listed in Table 2, where VIP is 1.25 or greater.

[0111] In any of the methods of the present invention, the multiple wavenumber ranges of the measured Raman spectrum may include four or more of the wavenumber ranges 34-40 with a VIP of 1.50 or greater, as listed in Table 2; or the multiple wavenumber ranges of the measured Raman spectrum may include five or six of the wavenumber ranges 34-40 with a VIP of 1.50 or greater, as listed in Table 2; or the multiple wavenumber ranges of the measured Raman spectrum may include all seven of the wavenumber ranges 34-40 with a VIP of 1.50 or greater, as listed in Table 2.

[0112] [Table 2]

[0113] The viral titer and / or viral component abundance can be measured using multiple wavenumber ranges in a Raman spectrum, as described above in relation to Table 2. In one preferred embodiment of the present invention, the viral titer and / or viral component abundance measured using multiple wavenumber ranges in a Raman spectrum, as described above in relation to Table 2, is the adeno-associated virus (AAV) titer. In one particularly preferred embodiment of the present invention, the viral titer and / or viral component abundance measured using multiple wavenumber ranges in a Raman spectrum, as described above in relation to Table 2, is the adeno-associated virus serotype 8 (AAV8) titer.

[0114] In any of the methods of the present invention, the multiple wavenumber ranges of the measured Raman spectrum may include five or more wavenumber ranges from 1 to 28, where VIP is 1.00 or greater, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include ten or more wavenumber ranges from 1 to 28, where VIP is 1.00 or greater, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include fifteen wavenumber ranges from 1 to 28, where VIP is 1.00 or greater, as listed in Table 3 below. It can include more than one. Multiple wavenumber ranges of the measured Raman spectrum can include 20 or more wavenumber ranges from 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3 below. Multiple wavenumber ranges of the measured Raman spectrum can include 25 or more wavenumber ranges from 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3 below. Multiple wavenumber ranges of the measured Raman spectrum can include all 28 wavenumber ranges from 1 to 28 with a VIP of 1.00 or greater, as listed in Table 3 below.

[0115] In any of the methods of the present invention, the multiple wavenumber ranges of the measured Raman spectrum may include five or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include ten or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include fifteen or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include twenty or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include twenty or more wavenumber ranges from 29 to 59 with a VIP of 1.25 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include 30 of the wavenumber ranges 29–59 where VIP is 1.25 or greater, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include all 31 of the wavenumber ranges 29–59 where VIP is 1.25 or greater, as listed in Table 3 below.

[0116] In any of the methods of the present invention, the multiple wavenumber ranges of the measured Raman spectrum may include five or more wavenumber ranges from 60 to 81 with a VIP of 1.50 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include ten or more wavenumber ranges from 60 to 81 with a VIP of 1.50 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include fifteen or more wavenumber ranges from 60 to 81 with a VIP of 1.50 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include twenty or more wavenumber ranges from 60 to 81 with a VIP of 1.50 or higher, as listed in Table 3 below. The multiple wavenumber ranges of the measured Raman spectrum may include all 22 wavenumber ranges from 60 to 81 with a VIP of 1.50 or higher, as listed in Table 3 below.

[0117] [Table 3]

[0118] The viral titer and / or viral component abundance can be measured using multiple wavenumber ranges in the Raman spectrum as described above with respect to Table 3. In a preferred embodiment of the present invention, the viral titer and / or viral component abundance measured using multiple wavenumber ranges in the Raman spectrum as described above with respect to Table 3 is the lentiviral titer and / or viral component abundance.

[0119] In another preferred embodiment of the present invention, a value greater than 1.5 (Variable Importance Index (VIP)) The peaks indicate wavenumber ranges that are likely to be important for determining the presence of the virus and the viral titer and / or viral component abundance. Determining the signal intensity in specific wavenumber ranges is essential for predicting viral titer and / or viral component abundance by the method of the present invention. The 1.5VIP threshold is used in embodiments of the present invention to determine which wavenumber ranges are important for predicting viral titer and / or viral component abundance. Thus, in a preferred embodiment, the present invention encompasses a method for evaluating or predicting viral titer and / or viral component abundance using Raman spectroscopy, comprising the step of determining the signal intensity in four or more of these wavenumber ranges from the Raman spectrum obtained from the sample. The measured preprocessed signals are used for prediction. Any subsequent spectra from which prediction is made (i.e., spectra after model construction) must be preprocessed using the exact same method as the data used to train and build the PLS model.

[0120] Any of the methods of the present invention includes the step of measuring the total intensity of Raman scattered light in each of a plurality of wavenumber ranges to obtain a wavenumber intensity dataset of the sample, wherein the plurality of wavenumber ranges are pre-selected and characteristic of the viral component in the sample.

[0121] The step of determining the signal intensity in each desired wavenumber range requires determining the level of any peak identified within the desired wavenumber range. As described above, the signal intensity in any wavenumber range that is thought to be related to viral titer and / or viral component abundance can be at any level, and those skilled in the art will understand that measuring such intensity, when analyzed with a suitable multivariate model, will enable the determination of viral titer and / or viral component abundance. Accordingly, as will be discussed in more detail below, the present invention may include the step of evaluating or calculating viral titer and / or viral component abundance by analyzing the signal intensity measured using a multivariate model. Such a multivariate model may be prepared before carrying out the present invention or as part of the method of the present invention, in which case the method may further include the step of constructing the multivariate model.

[0122] The method of the present invention may include determining the signal intensity in a further wavenumber range in addition to the wavelength range specified herein.

[0123] Neural networks can also be used for classification based on Raman spectra, for example, in the analysis of diseased and healthy tissues in pathology. Neural networks can also be used for regression problems, such as the problems encountered when applying Raman data for monitoring viral production, as described herein.

[0124] A more sophisticated approach utilizes what is called a Convolutional Neural Network (CNN) (deep learning), often using Google's TensorFlow backend and Keras API to write scripts in the object-oriented Python programming language. The advantage of using convolutional layers is that the need for preprocessing decreases as the network essentially "learns" the perfect way to preprocess the spectrum itself to predict the optimal titer / concentration. Such neural networks for use in the data processing steps described herein are well known to those skilled in the art. Further information can be found, for example, below. https: / / www.forbes.com / sites / bernardmarr / 2018 / 09 / 24 / what-are-artificial-neural-networks-a-simple-explanation-for-absolutely-anyone / #2c9442a91245 http: / / pages.cs.wisc.edu / ~bolo / shipyard / neural / local.html

[0125] As described above, multivariate data model parameters can be obtained and used in methods for quantifying viral titers and / or viral component abundances as defined herein. Such model parameters can also be used, if necessary, when constructing alternative multivariate data models as defined herein. The inventors applied a multivariate algorithm to Raman spectral wavenumber signal intensity data to obtain model parameters that are then used to quantify viral titers and / or viral component abundances. Specifically, the inventors applied a multivariate algorithm to obtain regression coefficients used as model parameters. However, those skilled in the art will understand that, depending on the nature of the selected model, alternative model parameters can be obtained and used. Thus, in any of the methods defined herein, multivariate data model parameters can be appropriately selected and may optionally include regression coefficients. Any suitable multivariate algorithm can be applied to Raman spectral wavenumber signal intensity data to obtain model parameters. Multivariate regression algorithms such as the Partial Least Squares (PLS) regression algorithm can be used, and optionally, the PLS algorithm is a Nonlinear Iterative Partial Least Squares (NIPALS) regression algorithm. Model parameters can also be obtained using algorithms that include neural networks.

[0126] Those skilled in the art will also understand that the method of the present invention may include additional mathematical data processing and modeling steps for quantifying the viral component of interest.

[0127] Model for measuring viral titer Chemometric modeling of Raman spectra was performed as described herein to identify correlations between real-time increases in viral titers and / or viral component abundances and the specificity and intensity of wavenumber ranges observed in the Raman spectra. The identified wavenumber ranges are described above.

[0128] As described above, the present invention requires the evaluation of signal intensity in wavenumber ranges of 4 or 5 or higher, which are determined to be important in the evaluation of viral titer and / or viral component abundance, as well as further analysis of the intensity using a constructed multivariate model (either calibrated or uncalibrated).

[0129] Those skilled in the art will understand that different multivariate models can be constructed depending on the sample being analyzed, and that methods for constructing multivariate models are well known in the art (see, for example, the references cited herein). Therefore, different multivariate models may be required, for example, to determine viral titers and / or viral component abundances in samples containing different types of viruses, different cell culture media, or different producer cells.

[0130] In one embodiment of the present invention, a multivariate model can be constructed using the following approach. i) As described above, in regression of pre-processed Raman data in offline responses obtained using qPCR and other methods such as p24 ELISA and plaque assays, the regression may include comparison of pre-processed Raman spectra. ii) Use the obtained regression coefficients to predict the response values ​​using the preprocessed data. The quality of these predictions can be optimized by adjusting the number of underlying components / factors used in the multivariate regression. iii) Perform variable selection using known methods, e.g., a Variable Importance Index (VIP) to identify strong / important variables for predicting Y, in addition to describing X. iv) In building a refined model, following the identification of key spectral variables, a further round of modeling can be performed using the same approach described in step (ii), but variables or columns in the preprocessed Raman spectrum array that are deemed irrelevant by variable selection are removed before the model is built. This results in a simpler model built on data from which much of the irrelevant variation has been removed. As in (ii), the number of underlying components can be optimized by selecting the model built with the fewest number of underlying factors, giving the smallest predictive error for variation between components. Using the regression coefficients determined in v)(iv), we make future predictions, where the new preprocessed Raman spectrum can be multiplied by the regression coefficients obtained to generate estimates.

[0131] Once the wavenumber range for the analysis is determined, it will become clear that it may not be necessary to repeat steps (i) to (iii) above. In particular, in this case, only step iv) may be required to build the model. Furthermore, data modeling parameters other than regression coefficients may be used.

[0132] In a further aspect of the present invention, the method may include an additional step of preparing or constructing a multivariate model.

[0133] As described in step v) above, the regression coefficients from the generated multivariate model can be used to obtain estimates of viral titers and / or viral component abundances from Raman spectra obtained from one or more samples. Therefore, in this invention, the method may include the step of determining viral titers and / or viral component abundances using regression coefficients from a multivariate model. The same preprocessing method can be used for training / building the model.

[0134] The present invention is further illustrated by the following embodiments, which should not be construed as further limitations. All figures and all references, patents, and published patent applications cited throughout this application are expressly incorporated herein by reference. [Examples]

[0135] Example 1 - Calculation of lentivirus concentration at various stages during the production process Calculations were performed to provide estimated lentivirus concentrations in mg / ml predicted at various stages of an exemplary virus production process. The concentrations were calculated using known suspension densities and physical titers of the lentivirus. The results are shown in the table below.

[0136] [Table 4]

[0137] As shown, the estimated lentivirus concentration was found to be in the range of 2.03E-02 to 2.03E-05 mg / ml.

[0138] For comparison, some useful calculations based on the detection limits of glucose and phenylalanine, obtained from Buckley and Ryder, 2017, Applied Spectroscopy, 71, pp. 1085-1116, are shown below.

[0139] glucose mw = 180.156 g / mol Detection limit 0.6 mM (see Buckley and Ryder) Mole = Volume / 1000 × Molar concentration 1 / 1000×0.0006-1ml = 0.0000006 moles Mass = moles × mw = 0.0000006 × 180.156 = 0.00011 g / ml Estimated concentration=0.11mg / ml

[0140] Phenylalanine mw = 165.19 g / mol LoD = 1.1 mM (see Buckley and Ryder paper) Mole = Volume / 1000 × Molar concentration 1 / 1000×0.0011-1ml = 0.0000011 moles Mass = moles × mw = 0.0000011 moles × 165.19 g / mol = 0.00018g Estimated concentration=0.18mg / ml

[0141] The detection limits for glucose and phenylalanine, measured in mg / ml, are 5 to 10 times higher than the optimistic estimates commensurate with their conservative physical titers.

[0142] Based on the above, the approximate concentration of lentivirus in mg / mL in the culture medium is predicted to be below the detection limit using Raman spectroscopy.

[0143] Example 2 - Lentiviral production from a transient HEK293 process Experimental method Cell culture and transient transfection HEK293 cultures were grown at 37°C in an Eppendorf DASbox BioBLU300 bioreactor in FreeStyle 293 expression medium (ThermoFisher) without additional supplements. Cells were agitated before transient transfection and grown for 2 days to produce lentiviruses. Cells were transfected with genomes encoding gag-pol, vsv-g, and eGFP, and LV particles were produced using Polyplus PEIPro.

[0144] Throughout the entire process, 10 or 12 samples were taken from each bioreactor to measure viral titer using qRT-PCR and confirmed by P24 ELISA. Raman spectra were obtained throughout the growth and viral production phases.

[0145] RT-qPCR PCR kit used: Lenti-X (trademark) qRT-PCR titration kit (manufactured by Takara, catalog number 631235) Supplier: Clontech Method of Operation: The kit is a one-step reverse transcription and PCR amplification kit. The primers in this kit target a conserved region of the HIV-1 genome adjacent to the packaging signal. The amplicon is detected by SYBR green fluorescence, and the final titer is determined from an ssRNA standard used to construct a standard curve. The final quantification of viral titer is provided as viral genome count / ml.

[0146] P24-ELISA ELISA Kit: QuickTiter (trademark) Lentivirus Titer Kit (Lentivirus-associated HIV p24) (Catalog Number VPK-107) Supplier: Cell Biolabs, Inc. Method of Action: The kit is an enzyme immunoassay developed for the detection and quantification of lentivirus-associated HIV-1 p24 core protein only. Virus-associated p24 can be quantified as p24 titer (ng / ml) or particle count / ml, assuming approximately 2000 p24 molecules per lentivirus particle.

[0147] Raman spectroscopy Raman measurements were performed using a Kasier Optics RxN2 Raman spectrometer, which has the capability to sequentially monitor four probe channels. The RxN2 excitation source was a 785 nm near-infrared diode laser with a nominal output of ~270 mW at each probe head. The samples consisted of the contents of four Eppendorf dasBox BioBLU disposable systems. The beam was delivered to each sample bioreactor using one set of four Kaiser Optics filtered fiber optic MR probes and a BioOptic 220 for each bioreactor. Prior to in-process measurements, the RxN2 system was stabilized for one hour, and then each of the four probe channels was calibrated using the RxN2's internal automatic calibration standard. Furthermore, CCD sensitivity correction was performed on each probe channel using a National Institute of Standards and Technology (NIST) certified light source (HCA). Scattered light was collected using the same BioOptic 220 and MR probes used for beam delivery. Within each MR probe, scattered light was transmitted via a second optical fiber to the RxN2 f\1.8 imaging spectrometer. After filtering the Rayleigh scattered light using a holographic notch filter, the Raman scattered light was directed to a Kaiser Optics holographic transmission diffraction grating and subsequently imaged onto a thermoelectrically cooled 1024-pixel CCD detector. The system's effective bandwidth was 100–3425 cm⁻¹. -1 The resolution is 4cm. -1 The Raman spectrum is obtained with an integration time of ~15 minutes / channel, including the CCD readout time, ranging from 100 to 3425 cm⁻¹. -1 Acquisitions were obtained from each bioreactor over 10 seconds, and averaged over 75 cumulative acquisitions to generate each measured spectrum. Each channel was measured sequentially. At various points throughout the process, liquid samples were acquired from each bioreactor, and those points were noted so that the offline assay data could be retrospectively matched to the corresponding Raman spectra.

[0148] Raman data analysis All data analysis was performed using MATLAB (The MathWorks, MA, USA) version R2017b. The raw Raman spectra were obtained by scanning the entire spectrum at ~3000 cm⁻¹. -1 The samples were pre-treated by normalizing to the peak intensity of the water band. A moderate fluorescence background signal was observed at 420–1800 cm⁻¹. -1 The region was removed. The lower limit of this range was chosen to avoid Raman bands that may arise from the sapphire window of the BioOptic-220 or that may be artifacts of the optical design of the Raman instrument and probe. Next, the reduced normalized spectrum was examined for obvious outliers and artifacts. Spectra relevant to the offline sampling time were identified, and a model training subset of the preprocessed spectra was prepared. Next, the training set of preprocessed spectra was used for chemometric modeling. Before chemometric modeling, the spectra were centered to the mean. Several initial projection-regression (PLS-R) models to latent structure for important analytes and viral titers were constructed. Using these models, multivariate Raman spectra can be regressiond on samples containing analytes of interest (viral titers) at known concentrations. Based on these calculations, the concentrations of analytes can be predicted in the future. The models were prepared using a 10-fold cross-validation procedure on the training data. That is, the data Tenths of the data were randomly selected and removed from the training data to be used to evaluate the model's performance. This was done 10 times, and the error values ​​and model accuracy / performance statistics are the mean values ​​obtained for each of the 10 training sets. Selecting the number of underlying components or basis vectors is a crucial step in building supervised linear models such as PLS-R. In this work, the optimal number of underlying components was identified by examining a plot of the mean squared prediction error (MSECV) after cross-validation as a function of the number of components. The minimum value identifies the optimal number of PLS ​​components. A second stage of variable selection is required to optimize the model built by selecting only the wavenumbers / variables most important for prediction. This was done using the Variable Importance Index (VIP) method. However, there are many ways to perform variable selection. A typical VIP plot is shown in Figure 5. Typically, variables with VIP values ​​greater than 1 are used in the final model. However, here we built and evaluated models using several VIP thresholds to identify the minimum number of spectral variables required to make good predictions and the threshold at which modeling of offline RT-qPCR data becomes impossible. As the VIP threshold increases, the number of spectral variables identified decreases. Once significant variables were determined for each VIP threshold, the final models were constructed. These models were then used to predict intermediate viral strength values, i.e., the viral strength values ​​between each offline data point for all available runs.

[0149] Preliminary evaluation of viral titer models and selection of scope Figure 2 shows an example of a pre-processed Raman spectrum used in chemometric modeling.

[0150] Viral titers were monitored throughout the entire project. Representative titers obtained by RT-qPCR are summarized in Figure 3.

[0151] Figure 4 shows a plot of the mean squared prediction error after cross-validation of the initial PLS-R model. It was found that the minimum effective prediction error occurred when using 12 PLS components, compared to using all spectral variables or channels.

[0152] Therefore, from this plot (Figure 4), we can conclude that the best compromise between minimizing prediction error and model simplicity lies in the 12-component model. Increasing the number of components may yield a slight improvement in predictive power, but this is likely simply a result of incorporating noise and overfitting the model to the training set. In other words, it predicts only the training data and not new, unseen data, building a model based on spurious correlations between measured variables and dependent / response variables.

[0153] Following the preparation of the initial 12-component model, various variable selection methods were evaluated to select the optimal / most predictive spectral variables for the final model. The goal here was to enhance its simplicity by removing unnecessary spectral channels / variables from the model and including only physically meaningful information. A Variable Importance Index (VIP) was ultimately calculated to determine which spectral variables were most important in predicting the virus copy number (Figure 5A). Several variable importance thresholds were investigated as criteria for retained variables to evaluate and identify the minimum number of spectral variables required to make an acceptable physical titer prediction. Typically, a VIP threshold of 1 was used, and thresholds from 1.00 to 1.75 were investigated. Figure 5B shows the range of variables or wavenumbers that the VIP algorithm identifies as being most important, i.e., greater than the selected threshold (1.5 in this case). Figure 5C shows these wavenumber ranges in order of importance.

[0154] After reducing the number of spectral variables, a further evaluation of the number of underlying latent variables was performed. The optimal number of PLS ​​components depends on the number of spectral variables or wavenumbers used in the model. This can vary. After reducing the spectral variables using a threshold of 1.5, we found that the optimal number of PLS ​​latent variables is 10. Figure 6 shows the MSECV plots of refined models with different numbers of underlying components. The fact that the mean squared prediction error increases with a larger number of underlying components indicates that if there are more than 10 PLS components, the resulting model is overfitting to the training set.

[0155] The model predictions for the RT-qPCR viral copy number for each of the three runs, estimated using regression coefficients obtained from a (conservative) model of 10 latent variables and selected spectral variables with VIP >= 1.5, are shown in Figures 7, 8, and 9 below, respectively. The results indicate that the model using Raman spectroscopy data is consistent with offline measurements of viral titers over time. A comparison of titers obtained from the RT-qPCR assay and P24 ELISA is shown in Figure 10.

[0156] Figure 11 illustrates a method for monitoring the stages of virus production culture using the method of the present invention. Because Raman spectroscopy is used continuously in real time, the method of the present invention can accurately track changes in the rate of virus production. Thus, changes from the start to the production stage and changes from the production stage to the end stage can be identified.

[0157] Example 3 - Evaluation and range selection of a sophisticated viral titer model To further refine the selection of wavenumber ranges for viral titer assessment, a further study was conducted using additional samples, employing the same methodology described above for Example 2. Using this approach, various wavenumber ranges for calculating viral titers were identified by applying a Variable Importance Index (VIP) threshold of 1.00 or higher. Additional wavenumber ranges were identified by applying a VIP threshold of 1.25 or higher, and further wavenumber ranges were identified by applying a VIP threshold of 1.50 or higher. The results are shown in Table 3 above.

[0158] Figure 12 shows a schematic diagram of the formula for quantifying viral titer. This formula is applied to regression coefficients obtained from a multivariate regression algorithm applied to normalized Raman signal intensity data.

[0159] Further analysis was performed to determine the number of wavenumber ranges that can be used to provide an accurate estimate of viral titer.

[0160] The range identified as important for viral vector production, i.e., the range identified as important by a Variable Importance Index (VIP) of 1.00 or higher after initial PLS modeling using the extended spectral range (~420~1800 cm-1) (i.e., wavenumber ranges 1~28 listed in Table 3 above), was identified and further analysis was performed.

[0161] The data was divided into randomly selected training and test data pair blocks in a 4:1 ratio, i.e., Raman spectra and their associated offline viral titer data for model building (80%) and model testing (20%).

[0162] Various combinations of ranges considered important by VIP are probabilistically evaluated for various training and test pairs, i.e., for a total of r values ​​for each range from 1 to 28, and the model performance R 2 Statistics (Note: R 2 Many combinations were evaluated based on (=1 - sum of squared residuals / total sum of squares), and the standard deviation of the performance of various models was evaluated to generate confidence intervals. The minimum number of ranges was the mean R of several training / test pairs of data. 2 This was determined by selecting a number in the range where the mean of the values ​​is approximately 0.5. Figure 13 shows R as a function of the number in the wavenumber range. 2 This shows the plot.

[0163] This analysis identified 5 as the minimum wavenumber range required to provide an estimate of the viral titer.

[0164] Therefore, in any of the methods of the present invention, the viral titer can be calculated using five or more wavenumber ranges from 1 to 28 as shown in Table 3, specified by a VIP threshold of 1.00 or higher, as described in more detail herein. In any of the methods of the present invention, preferably, as described in more detail herein, the viral titer can be calculated using five or more wavenumber ranges from 29 to 59 as shown in Table 3, specified by a VIP threshold of 1.25 or higher, as described in more detail herein. In any of the methods of the present invention, more preferably, as described in more detail herein, the viral titer can be calculated using five or more wavenumber ranges from 60 to 81 as shown in Table 3, specified by a VIP threshold of 1.50 or higher, as described in more detail herein. In any of these methods, preferably, as described in more detail herein, the viral titer can be calculated using 10 or more wavenumber ranges, more preferably 15 or more, and even more preferably 20 or more.

[0165] Example 4 - AAV8 production from a transient HEK293 process Experimental method Cell culture and transient transfection HEK293F cultures were grown in an Eppendorf DASbox BioBLU300 bioreactor at 37°C in BalanCD medium (Irvine Scientific) containing 4 mM GlutaMAX (Fisher). Cells were agitated before transient transfection and grown for 24 hours to produce AAV8. Cells were transfected with serum-free Opti-MEM (Gibco) with rep, cap, genome-encoded eGFP plasmids, and helper plasmids (E2A, E4), and AAV8 particles were produced using PEIPro (Polyplus transfection).

[0166] Throughout the entire process, 12 samples were taken from each bioreactor to measure viral titer using qRT-PCR. Raman spectra were obtained throughout the proliferation and viral production phases.

[0167] RT-qPCR Viral titers of AAV8 samples were measured using TaqMan®-based real-time qPCR, and the final quantification was provided as viral genome count / mL (VG / mL). Assay primers targeted the ITR2 sequence of the AAV8 viral genome. Amplicons were detected by TaqMan® fluorescence-generating probes. Viral titers were determined from standard curves constructed from linearized plasmids.

[0168] Raman spectroscopy Raman measurements were performed using a Kasier Optics RxN2 Raman spectrometer. This spectrometer has the capability to sequentially monitor four probe channels. The RxN2 excitation source was a 785 nm near-infrared diode laser with a nominal output of ~270 mW at each probe head. The samples consisted of the contents of four Eppendorf dasBox BioBLU disposable systems. The beam was delivered to each sample bioreactor using one set of four Kaiser Optics filtered fiber optic MR probes and a BioOptic220 for each bioreactor. Before in-process measurements, the RxN2 system was stabilized for 1 hour, and then each of the four probe channels was calibrated using the RxN2's internal automatic calibration standard, and further CCD sensitivity correction was performed on each probe channel using a National Institute of Standards and Technology (NIST) certified light source (HCA). The experiment was performed using a 3D fiber optic cable. Scattered light was collected using the same BioOptic 220 and MR probes used for beam delivery. Within each MR probe, scattered light was delivered via a second optical fiber to the RxN2 f / 1.8 imaging spectrometer. After filtering the Rayleigh scattered light using a holographic notch filter, the Raman scattered light was directed to a Kaiser Optics holographic transmission diffraction grating and subsequently imaged onto a thermoelectrically cooled 1024-pixel CCD detector. The system's effective bandwidth was 100–3425 cm⁻¹. -1 The resolution is 4cm.-1 The Raman spectrum is obtained with an integration time of ~15 minutes / channel, including the CCD readout time, ranging from 100 to 3425 cm⁻¹. -1 Acquisitions were obtained from each bioreactor over 10 seconds, and averaged over 75 cumulative acquisitions to generate each measured spectrum. Each channel was measured sequentially. At various points throughout the process, liquid samples were acquired from each bioreactor, and those points were noted so that the offline assay data could be retrospectively matched to the corresponding Raman spectra.

[0169] Raman data analysis All data analysis was performed using MATLAB (The MathWorks, MA, USA) version R2019b. The raw Raman spectra were obtained by scanning the entire spectrum at ~3000 cm⁻¹. -1 The samples were pre-treated by normalizing to the peak intensity of the water band. A moderate fluorescence background signal was observed at 420–1800 cm⁻¹. -1The region was removed. The lower limit of this range was chosen to avoid Raman bands that may arise from the sapphire window of the BioOptic-220 or that may be artifacts of the optical design of the Raman instrument and probe. Next, the reduced normalized spectrum was examined for obvious outliers and artifacts. Spectra relevant to the offline sampling time were identified, and a model training subset of the preprocessed spectra was prepared. Next, the training set of preprocessed spectra was used for chemometric modeling. Before chemometric modeling, the spectra were centered to the mean. Several initial projection-regression (PLS-R) models to latent structure for important analytes and viral titers were constructed. These models allow multivariate Raman spectra to be regressiond against samples containing interesting analytes (viral titers) at known concentrations. Based on these calculations, the concentrations of analytes can be predicted in the future. The models were prepared using a 10-fold cross-validation procedure on the training data. That is, 1 / 10 of the data was randomly selected and removed from the training data and used to evaluate the performance of the models. This was performed 10 times, and the error values ​​and model accuracy / performance statistics are the mean values ​​obtained for each of the 10 training sets. Choosing the number of underlying components or basis vectors is a crucial step in building supervised linear models such as PLS-R. In this work, the optimal number of underlying components was identified by examining a plot of the mean squared prediction error (MSECV) after cross-validation as a function of the number of components. The minimum value identifies the optimal number of PLS ​​components. A second stage of variable selection is required to optimize the model built by selecting only the wavenumbers / variables most important for prediction. This was done using the Variable Importance Index (VIP) method. However, there are many ways to perform variable selection. A typical VIP plot is shown in Figure 18. Typically, variables with VIP values ​​greater than 1 are used in the final model.However, here we built and evaluated models using several VIP thresholds to identify the minimum number of spectral variables required to make good predictions and the thresholds at which modeling of offline RT-qPCR data becomes impossible. As the VIP threshold increases, the number of spectral variables identified decreases. Once significant variables were determined for each VIP threshold, the final models were built. These models were then used to predict intermediate viral strength values, i.e., the viral strength values ​​between each offline data point of all available runs.

[0170] Preliminary evaluation of viral titer models and selection of scope Exemplary pre-processed Lamans when used for AAV chemometric modeling. The vector is shown in Figure 14.

[0171] AAV titers were monitored throughout the entire project, and representative titers obtained by RT-qPCR are summarized in Figure 15.

[0172] Figure (16) shows a plot of the mean squared prediction error after cross-validation of the initial PLS-R model. It was found that the smallest effective prediction error occurred when using 15 PLS components, compared to using all spectral variables or channels.

[0173] Therefore, from this plot (Figure 16), we can conclude that the best compromise between minimizing prediction error and model simplicity lies in the 15-component model. Following the preparation of the initial 15-component model, various variable selection methods were evaluated to select the optimal / most predictive spectral variables for the final model. The goal here was to enhance its simplicity by removing unnecessary spectral channels / variables from the model and including only physically meaningful information. A Variable Importance Index (VIP) was ultimately calculated to determine which spectral variables were most important in predicting the virus copy number (Figure 17A). Several variable importance thresholds were investigated as criteria for retained variables to evaluate and identify the minimum number of spectral variables required to make an acceptable physical titer prediction. Typically, a VIP threshold of 1 was used, and thresholds from 1.00 to 1.75 were investigated. Figure 17B shows the range of variables or wavenumbers that the VIP algorithm identifies as the region considered most important, i.e., the region greater than the selected threshold (1.0 in this case). Figure 17C shows these wavenumber ranges in order of importance.

[0174] After reducing the number of spectral variables, a further evaluation of the number of underlying latent variables was performed. The optimal number of PLS ​​components may vary depending on the number of spectral variables or wavenumbers used in the model. After reducing the spectral variables using a threshold of 1.0, it was found that the optimal number of PLS ​​latent variables is 9. Figure 18 shows the MSECV plots of refined models with different numbers of underlying components. The fact that the mean squared prediction error increases with a larger number of underlying components indicates that if there are more than 9 PLS components, the resulting model is overfitted to the training set.

[0175] Figure 19 below shows model predictions of RT-qPCR viral copy numbers for four exemplary bioreactor runs, estimated using regression coefficients obtained from a (conservative) model of nine latent variables and selected spectral variables with VIP > 1.0. The results indicate that the model using Raman spectroscopy data is consistent with offline measurements of viral titers over time.

[0176] Example 5 - Evaluation and range selection of a sophisticated viral titer model for AAV Further analysis was performed on the sample described in Example 4 to analyze the number of wavenumber ranges that can be used to provide an accurate estimate of AAV virus titer.

[0177] The range identified as important for viral vector production, i.e., the range identified as important by a Variable Importance Index (VIP) of 1.00 or higher after initial PLS modeling using the extended spectral range (~420~1800 cm-1) (i.e., wavenumber ranges 1-12 listed in Table 1 above), was identified and further analysis was performed.

[0178] The data was divided into randomly selected training and test data pair blocks in a 4:1 ratio, i.e., Raman spectra and their associated offline viral titer data for model building (80%) and model testing (20%).

[0179] Various combinations of ranges considered important by VIP are probabilistically evaluated for various training and test pairs, i.e., for a total of r values ​​for each range from 1 to 12, and the model's performance R 2 Statistics (Note: R 2 Many combinations were evaluated based on (=1 - sum of squared residuals / total sum of squares), and the standard deviation of the performance of various models was evaluated to generate confidence intervals. The minimum number of ranges was the mean R of several training / test pairs of data. 2 This was determined by selecting a number in the range where the mean of the values ​​is approximately 0.5. Figure 20 shows R as a function of the number in the wavenumber range. 2 This shows the plot.

[0180] This analysis identified 4 as the minimum wavenumber range required to provide an estimate of AAV virus titer.

[0181] Therefore, in any of the methods of the present invention, the viral titer can be calculated using four or more wavenumber ranges 1 to 12 as shown in Table 1, which are specified by a VIP threshold of 1.00 or higher, as described in more detail herein. In any of these methods, the viral titer can be calculated using preferably six or more wavenumber ranges, more preferably eight or more, even more preferably ten or more, or most preferably all twelve, as described in more detail herein. In any of the methods of the present invention, the viral titer can be calculated using four or more wavenumber ranges 13 to 22 as shown in Table 1, which are specified by a VIP threshold of 1.25 or higher, as described in more detail herein. In any of these methods, the viral titer can be calculated using preferably six or more wavenumber ranges, more preferably eight or more, or most preferably all ten, as described in more detail herein. As described in more detail herein, the viral titer can be calculated using four or more wavenumber ranges 23 to 30 as shown in Table 1, which are specified by a VIP threshold of 1.50 or higher, as described in more detail herein. In any of these methods, the viral titer can be calculated using preferably six or more wavenumber ranges, or most preferably all eight, as described in more detail herein.

[0182] Example 7 - AAV8 production from a transient HEK293 process and determination of the Empty-Full ratio. Experimental method Cell culture and transient transfection HEK293F cultures were grown in an Eppendorf DASbox BioBLU 300 bioreactor at 37°C in BalanCD medium (Irvine Scientific) containing 4 mM GlutaMAX (Fisher). Cells were agitated before transient transfection and grown for 24 hours to produce AAV8. Cells were transfected with serum-free Opti-MEM (Gibco) with rep, cap, genomically coded eGFP plasmid, and helper plasmids (E2A, E4), and AAV8 particles were produced using PEIPro (Polyplus transfection).

[0183] Throughout the process, 11 samples were obtained from each bioreactor, and viral titers were measured using RT-qPCR (genome copy number per ml). Five of these samples were further used for ELISA (total particle count per ml). Raman spectra were obtained throughout the growth and viral production phases.

[0184] RT-qPCR Viral titers of AAV8 samples were measured using TaqMan®-based real-time qPCR, and the final quantification was provided as viral genome count / mL (VG / mL). Assay primers targeted the ITR2 sequence of the AAV8 viral genome. Amplicons were detected by TaqMan® fluorescent probes. Viral titers were determined from standard curves constructed from linearized plasmids.

[0185] ELISA Total AAV8 capsid titers were determined in extracellular AAV8 samples by ELISA, and the final quantification was provided as total particles / mL (TP / mL). To accurately quantify TP / mL for each sample, standard curves were created using reconfigured AAV8 standards with known particle concentrations. To perform the ELISA, mouse monoclonal antibodies specific to the conformational epitope on assembled AAV8 capsids (clone ADK8) were coated onto strips of microtiter plates and used to capture AAV8 particles in the samples. Captured AAV8 particles were detected using two steps: 1) conjugation of biotin-conjugated anti-AAV8 antibody to an immune complex, and 2) reaction of streptavidin peroxidase conjugate with biotin molecules. Upon addition of tetramethylbenzidine (TMB) substrate solution, a color reaction occurred proportional to the amount of specifically bound viral particles. Absorbance was then photometrically measured at 450 nm.

[0186] Raman spectroscopy Raman measurements were performed using a Kaiser Optics RxN2 Raman spectrometer. This spectrometer has the capability to continuously monitor four probe channels. The RxN2 excitation source was a 785 nm near-infrared diode laser with a nominal output of ~270 mW at each probe head. The sample consisted of the contents of four Eppendorf dasBox BioBLU disposable systems. The beam was from four Kaiser Optics probes. One set of Kaiser Optics filtered fiber optic MR probes and a BioOptic 220 was used for each bioreactor, delivering the light to each sample bioreactor. Prior to in-process measurements, the RxN2 system was stabilized for 1 hour, after which each of the four probe channels was calibrated using the RxN2's internal automatic calibration standard. Furthermore, CCD sensitivity correction was performed on each probe channel using a National Institute of Standards and Technology (NIST) certified hazardous material (HCA). Scattered light was collected using the same BioOptic 220 and MR probes used for beam delivery. Within each MR probe, scattered light was delivered via a second fiber optic to the RxN2 f / 1.8 imaging spectrometer. After filtering the Rayleigh scattered light using a holographic notch filter, the Raman scattered light was directed to a Kaiser Optics holographic transmission grating and subsequently imaged onto a thermoelectrically cooled 1024-pixel CCD detector. The system's effective bandwidth was 100–3425 cm⁻¹. -1 The resolution is 4cm. -1 The Raman spectrum is obtained at an integration time of approximately 15 minutes / channel, including the CCD readout time, with values ​​ranging from 100 to 3425 cm⁻¹. -1 Acquisitions were obtained from each bioreactor over 10 seconds, and averaged over 75 cumulative acquisitions to generate each measured spectrum. Each channel was measured sequentially. At various points throughout the process, liquid samples were acquired from each bioreactor, and those points were noted so that the offline assay data could be retrospectively matched to the corresponding Raman spectra.

[0187] Raman data analysis All data analysis was performed using MATLAB (The MathWorks, MA, USA) version R2019b. The raw Raman spectra were obtained by scanning the entire spectrum at ~3000 cm⁻¹. -1 The samples were pre-treated by normalizing to the peak intensity of the water band. A moderate fluorescence background signal was observed at 420–1800 cm⁻¹. -1The region was removed. The lower limit of this range was chosen to avoid Raman bands that may originate from the sapphire window of the BioOptic-220 or that may be artifacts of the optical design of the Raman instrument and probe. Next, the reduced normalized spectrum was examined for obvious outliers and artifacts. Spectra relevant to the offline sampling time were identified, and a model training subset of the preprocessed spectrum was prepared. Next, the training set of the preprocessed spectrum was used for chemometric modeling. Used in Delling. Before chemometric modeling, the spectrum was centered at the mean. Several initial projection-regression (PLS-R) models to latent structure for key analytes and viral titers (one model based on RT-qPCR calibrated to genome copy number per ml and AAV8 calibrated to total particle number per ml) were used. A single viral titer model based on ELISA was constructed. These models allow for regression of multivariate Raman spectra against samples containing analytes of interest at known concentrations (in this example, viral titers determined from various RT-qPCR and ELISA assays). Based on these calculations, the concentration of the analyte can be predicted in the future. The models were prepared using a 10-fold cross-validation procedure on the training data. That is, 1 / 10 of the data was randomly selected, removed from the training data, and used to evaluate the model's performance. This was done 10 times, and the error value, model precision / performance statistics, are the mean values ​​obtained for each of the 10x training sets. Choosing the number of underlying components or basis vectors is a crucial step in constructing supervised linear models such as PLS-R. In this work, the optimal number of underlying components was identified for each model by examining a plot of the mean squared predictive error (MSECV) after cross-validation as a function of the number of components. The minimum value identifies the optimal number of PLS ​​components for a given model. A second stage of variable selection is required to optimize the model, which was built by selecting only the wavenumbers / variables most important for prediction. This was done using the Variable Importance Index (VIP) method. However, many methods exist for performing variable selection. A typical VIP plot is shown in Figure 26A. Typically, variables with a VIP value greater than 1 are used in the final model. However, here we built and evaluated models using several VIP thresholds to identify the minimum number of spectral variables required to make good predictions and the thresholds at which it becomes impossible to model the offline data of interest. As the VIP threshold increases, the number of spectral variables identified decreases. Once significant variables were determined for each VIP threshold, the final models were built. These models were then used to predict intermediate viral force values ​​(i.e., viral force values ​​between each offline data point of all available runs) as both genome copy number per ml and total particle number per ml from Raman spectra.

[0188] Preliminary evaluation of viral titer models and selection of scope Figure 21 shows an example of a pre-treated Raman spectrum used for chemometric modeling of AAV titers, both in terms of genome copy number per ml and total particle count per ml.

[0189] AAV titers were monitored throughout the project. Figure 22 summarizes representative titers (genome copy number per ml) obtained by RT-qPCR, and Figure 23 shows representative total particle counts per ml obtained by ELISA.

[0190] Figure 24 shows a plot of the mean squared prediction error of the initial PLS-R model after cross-validation for genome copy number per ml. It was found that the minimum effective prediction error occurred when using 15 PLS components, with all spectral variables or channels being used. Another plot of the mean squared prediction error of the initial PLS-R model after cross-validation for total particle number per ml, calibrated from ELISA data, is shown in Figure 25. It was found that the minimum effective prediction error occurred when using 14 PLS components, with all spectral variables or channels being used. These selections of component numbers provide a good compromise between minimizing prediction error and model simplicity.

[0191] Following the preparation of initial 15- and 14-component models, various variable selection methods were evaluated to select the optimal / most predictive spectral variables for the final RT-qPCR and ELISA calibration models, respectively. The objective here was to remove unwanted spectral channels / variables from the two models. The goal was to simplify the variables by removing numbers and including only physically meaningful information. A Variable Importance Index (VIP) was ultimately calculated to determine which spectral variables were most important in predicting the viral copy number (Figure 26A). Several variable importance thresholds were investigated as criteria for retained variables to evaluate and identify the minimum number of spectral variables required to make an acceptable physical titer prediction. Typically, a VIP threshold of 1 was used, and thresholds from 1.00 to 1.75 were investigated. Figure 26B shows the range of variables or wavenumbers that the VIP algorithm identifies as being most important for predicting the genome copy number per ml, i.e., greater than the selected threshold (1.0 in this case). Figure 26C shows these wavenumber ranges in order of importance.

[0192] Furthermore, a similar analysis was performed using the above ELISA to identify the most important spectral variables for predicting the number of particles per ml. A Variable Importance Index (VIP) was calculated to determine which spectral variables were most important in predicting the number of viral particles (Figure 27A). Several variable importance thresholds were investigated as criteria for retained variables to evaluate and identify the minimum number of spectral variables required to make an acceptable physical titer prediction. Typically, a VIP threshold of 1 was used, and thresholds from 1.00 to 1.75 were investigated. Figure 27B shows the range of variables or wavenumbers that the VIP algorithm identifies as being most important, i.e., greater than the selected threshold (1.0 in this case). Figure 27C shows these wavenumber ranges in order of importance.

[0193] After reducing the number of spectral variables, further evaluation of the number of underlying latent variables was performed for both the viral copy number model and the viral particle number model, respectively. The optimal number of PLS ​​components may differ depending on the number of spectral variables or wavenumbers used in the model. After reducing the spectral variables using a threshold of 1.0, the optimal number of PLS ​​latent variables was found to be 10 for both the genome copy number per 1 ml of model and the particle number per 1 ml of model. Figures 28 and 29 show MSECV plots of refined models with different numbers of underlying components.

[0194] Figures 30 and 31 below show model predictions of RT-qPCR viral copy numbers for exemplary runs of eight bioreactors, estimated using regression coefficients obtained from a (conservative) model of 10 latent variables and selected spectral variables with VIP > 1.0. The results show that the model using Raman spectroscopy data is consistent with offline measurements of viral titer (genome copies per ml) over time. Figures 32 and 33 below show similar predictions from total ELISA particle count for exemplary runs of eight bioreactors, estimated using regression coefficients obtained from a (conservative) model of 10 latent variables and selected spectral variables with VIP > 1.0. The results show that the model using Raman spectroscopy data is consistent with offline measurements of viral titer (particles per ml) over time. A method for estimating the Empty-Full-Ratio of individual AAV samples as a percentage is to divide the number of genome copies per ml (RT-qPCR) by the total number of particles per ml (ELISA) and multiply this number by 100. Based on both of these methods for determining viral titer, similar calculations can be performed using the output from the predictive models developed above (Figures 30-33). The results of these calculations are shown in Figures 34 and 35, and as can be seen from these transfected cultures, the Empty-Full-Ratio is in good agreement with the estimates made using offline RT-qPCR and ELISA data.

[0195] Example 8 - Evaluation and range selection of a sophisticated ELISA virus titer model for AAV Further analysis was performed on the AAV8 ELISA model training, such as those described in Examples 3 and 5 above, to determine the AAV virus titer (in particular, the total number of particles per ml). We were able to calculate the number of wavenumber ranges necessary to provide an estimate.

[0196] The range identified as important for AAV8 ELISA, i.e., the range identified as important by a Variable Importance Index (VIP) of 1.00 or higher after initial PLS modeling using the extended spectral range (~420~1800 cm⁻¹) (i.e., wavenumber range 1–20 shown in Figures 27B and 27C above), was used for further analysis.

[0197] The data was divided into randomly selected training and test data pair blocks in a 4:1 ratio, i.e., Raman spectra and their associated offline viral titer data for model building (80%) and model testing (20%).

[0198] Various combinations of ranges considered important by VIP are probabilistically evaluated for various training and test pairs, i.e., for a total of r values ​​in each range from 1 to 20, and the performance statistics of the model (e.g., R) are calculated. 2 Statistics (Note: R 2 Many combinations were evaluated based on (1 - sum of squared residuals / total sum of squares), and the standard deviation of the performance of various models was evaluated to generate confidence intervals. The minimum number of ranges was the mean R of several training / test pairs of data. 2 This was determined by selecting a number in the range where the mean of the values ​​is approximately 0.5. 2 Other performance statistics can also be used with this approach.

[0199] References ·Barnes, R. J., Dhanoa, M. S. and Lister, S. J. Standard Normal Variate Transformation and De-trending of Near-Infrared Diffuse Reflectance Spectra ("Standard Normal Variate Transformation and De-trending of Near-Infrared Diffuse Reflectance Spectra"). Applied Spectroscopy. 1989, 43(5): pp772-777. ·Buckley, K. and Ryder, A. G. Applications of Raman Spectroscopy in Biopharmaceutical Manufacturing ("Applications of Raman Spectroscopy in Biopharmaceutical Manufacturing"): A Short Review. Applied Spectroscopy. 2017, 71(6): pp1085-1116. ·Hu H., Bai, J., Xia, G., Zhang, W., Ma, Y. Improved Baseline Correction Method Based on Polynomial Fitting for Raman Spectroscopy ("Improved Baseline Correction Method Based on Polynomial Fitting for Raman Spectroscopy"). Photonic Sensors. 2018, 8(4): pp332-340. ·Huang, J., Romero-Torres, S. and Moshgbar, M. Practical Considerations in Data Pre-treatment for NIR and Raman Spectroscopy ("Practical Considerations in Data Pre-treatment for NIR and Raman Spectroscopy"). American Pharmaceutical Review. 2010. ·Koch, M., Suhr, C., Roth, B. and Meinhardt-Wollweber, M. Iterative morphological and mollifier-based baseline correction for Raman spectra ("Iterative Morphological and Mollifier-based Baseline Correction for Raman Spectra"). Journal of Raman Spectroscopy. 2017, 48(2): pp336-342. ·Lee, J. H., Kim, B. C., Oh, B. K. and Choi, J. W. Rapid and Sensitive Determination of HIV-1 Virus Based on Surface Enhanced Raman Spectroscopy ("Rapid and Sensitive Determination of HIV-1 Virus Based on Surface Enhanced Raman Spectroscopy"). J. Biomed. Nanotechnol. 2015, 11(12): pp2223-2230. ·Lieber, C. A. and Mahadevan-Jansen, A. Automated method for subtraction of fluorescence from biological Raman spectra ("An Automated Method for Fluorescence Subtraction from Biological Raman Spectra"). Applied Spectroscopy. 2003, 57(11): pp1363-1367. ·Savitzky, A. and Golay, M. J. E. Smoothing and Differentiation of Data by Simplified Least Squares Procedures ("Smoothing and Differentiation of Data by Simplified Least Squares Procedures"). Analytical Chemistry. 1964, 36 (8): pp1627-1639. ·Wold, S., Sjostrom, M and Eriksson, L. PLS-regression: a basic tool of chemometrics. Chemometrics and Intelligent Laboratory Systems. 2001, 58: pp109-130. ·Zhang, ZM, Chen, S. and Liang, YZ Baseline correction using adaptive iteratively reweighted penalized least squares. Analyst. 2010, 135(5): pp1138-1146. • Zhao, J., Lui, H., McLean, DI and Zeng, H. Automated autofluorescence background subtraction algorithm for biomedical Raman spectroscopy. Applied Spectroscopy. 2007, 61(11): pp1225-32.

Claims

1. A method for determining the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in a sample using Raman spectroscopy, (a) Providing a sample and irradiating the sample with a light source, (b) (i) A step of measuring the total intensity of Raman scattered light in each of a first plurality of wavenumber ranges in order to obtain a first wavenumber intensity dataset of the sample, wherein the first plurality of wavenumber ranges are pre-selected and characteristic of the viral nucleic acid in the sample. (ii) The step of performing a first set of mathematical data processing steps on the first wavenumber intensity dataset, (iii) A step of determining the viral nucleic acid content of the sample based on the output of the first set of mathematical data processing steps, (c) (i) A step of measuring the total intensity of Raman scattered light in each of a second plurality of wavenumber ranges in order to obtain a second wavenumber intensity dataset of the sample, wherein the second plurality of wavenumber ranges are pre-selected and characteristic of one or more viral structural molecules of the virus in the sample. (ii) The step of performing a second set of mathematical data processing steps on the second wavenumber intensity dataset, (iii) A step of determining the virus content in the sample, which includes the one or more viral structural molecules, based on the output of the second set of mathematical data processing steps, (d) The step of determining the ratio of viral nucleic acid to a virus containing one or more viral structural molecules in the sample, based on the values ​​determined in steps (b)(iii) and (c)(iii), method.

2. The step of performing the first and second sets of mathematical data processing steps on the first and second wavenumber intensity datasets is: (i) optionally, a step of normalizing the wavenumber signal intensity data, which is performed by preprocessing the signal intensity data using one or more preprocessing analysis methods such as first derivative, second derivative, standard normal variable (SNV) method, polynomial fitting, multi-polynomial fitting, Morifier method, piecewise polynomial fitting (PPF) method, or adaptive iteratively reweighted penalty least squares (airPLS) method. (ii) A step of obtaining model parameters by applying a multivariate regression algorithm, such as a partial least squares (PLS) regression algorithm, to the wavenumber signal intensity data, wherein the PLS algorithm is optionally a nonlinear iterative partial least squares (NIPALS) regression algorithm or a neural network, (iii) The step of determining the viral nucleic acid content of the sample using the model parameters obtained by applying the multivariate regression algorithm to the signal intensity data, and determining the viral content of the sample, which includes one or more viral structural molecules: The method according to claim 1.

3. The method according to claim 1 or 2, wherein the light source used to irradiate the sample is a laser, and the sample is irradiated with light having a wavelength of 785 nm.

4. The method according to any one of the preceding claims, wherein the Raman scattered light is detected using a charge-coupled device (CCD).

5. The first set of wavenumber ranges of the Raman spectrum measured to obtain the first wavenumber intensity dataset of the sample includes four or more wavenumber ranges 1 to 12, as listed in Table 1, where VIP is 1.00 or greater, or The multiple wavenumber ranges of the measured Raman spectrum include six or more wavenumber ranges 1 to 12, as listed in Table 1, where VIP is 1.00 or greater, or The multiple wavenumber ranges of the measured Raman spectrum include eight or more of the wavenumber ranges 1 to 12, where VIP is 1.00 or greater, as listed in Table 1, or The multiple wavenumber ranges of the measured Raman spectrum include 10 or more wavenumber ranges 1 to 12, as listed in Table 1, where VIP is 1.00 or greater, or The multiple wavenumber ranges of the measured Raman spectrum include all 12 of the wavenumber ranges 1 to 12, where VIP is 1.00 or greater, as listed in Table 1. The method according to any one of claims 1 to 4.

6. The first set of wavenumber ranges of the Raman spectrum measured to obtain the first wavenumber intensity dataset of the sample includes four or more of the wavenumber ranges 13 to 22, where VIP is 1.25 or greater, as listed in Table 1, or The multiple wavenumber ranges of the measured Raman spectrum include six or more wavenumber ranges from 13 to 22, where VIP is 1.25 or greater, as listed in Table 1, or The multiple wavenumber ranges of the measured Raman spectrum include eight or more of the wavenumber ranges 13 to 22, where VIP is 1.25 or greater, as listed in Table 1, or The multiple wavenumber ranges of the measured Raman spectrum include all 10 of the wavenumber ranges 13 to 22, where VIP is 1.25 or greater, as listed in Table 1. The method according to any one of claims 1 to 4.

7. The first set of wavenumber ranges of the Raman spectrum measured to obtain the first wavenumber intensity dataset of the sample includes four or more wavenumber ranges 23 to 30, where VIP is 1.50 or greater, as listed in Table 1, or The multiple wavenumber ranges of the measured Raman spectrum include six or more wavenumber ranges from 23 to 30, where VIP is 1.50 or greater, as listed in Table 1, or The multiple wavenumber ranges of the measured Raman spectrum include all eight of the wavenumber ranges 23 to 30, where VIP is 1.50 or greater, as listed in Table 1. The method according to any one of claims 1 to 4.

8. The second set of wavenumber ranges of the Raman spectrum measured to obtain the second wavenumber intensity dataset of the sample includes four or more wavenumber ranges from 1 to 20, where VIP is 1.00 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include six or more wavenumber ranges from 1 to 20, where VIP is 1.00 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include eight or more wavenumber ranges from 1 to 20, where VIP is 1.00 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include 10 or more wavenumber ranges from 1 to 20, where VIP is 1.00 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include 12 or more, 14 or more, 16 or more, or 18 or more wavenumber ranges 1 to 20, where VIP is 1.00 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include all 20 of the wavenumber ranges 1 to 20, where VIP is 1.00 or greater, as listed in Table 2. The method according to any one of claims 1 to 7.

9. The second set of wavenumber ranges of the Raman spectrum measured to obtain the second wavenumber intensity dataset of the sample includes four or more of the wavenumber ranges 21 to 33, as listed in Table 2, where VIP is 1.25 or greater, or The multiple wavenumber ranges of the measured Raman spectrum include six or more of the wavenumber ranges 21 to 33, where VIP is 1.25 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include eight or more of the wavenumber ranges 21 to 33, where VIP is 1.25 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include 10 or more, 11 or more, or 12 of the wavenumber ranges 21 to 33, where VIP is 1.25 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include all 13 of the wavenumber ranges 21 to 33, where VIP is 1.25 or greater, as listed in Table 2. The method according to any one of claims 1 to 7.

10. The second set of wavenumber ranges of the Raman spectrum measured to obtain the second wavenumber intensity dataset of the sample includes four or more of the wavenumber ranges 34-40, where VIP is 1.50 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include five or six of the wavenumber ranges 34 to 40, where VIP is 1.50 or greater, as listed in Table 2, or The multiple wavenumber ranges of the measured Raman spectrum include all seven of the wavenumber ranges 34-40, where VIP is 1.50 or greater, as listed in Table 2. The method according to any one of claims 1 to 7.

11. The method according to any one of the preceding claims, wherein the virus in the sample is adeno-associated virus (AAV).

12. The first set of wavenumber ranges of the Raman spectrum measured to obtain the first wavenumber intensity dataset of the sample includes five or more wavenumber ranges from 1 to 28, as listed in Table 3, with a Variable Importance Index (VIP) of 1.00 or higher, or The multiple wavenumber ranges of the measured Raman spectrum include 10 or more wavenumber ranges from 1 to 28, where VIP is 1.00 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 15 or more wavenumber ranges from 1 to 28, where VIP is 1.00 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 20 or more wavenumber ranges from 1 to 28, where VIP is 1.00 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 25 or more wavenumber ranges from 1 to 28, where VIP is 1.00 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include all 28 of the wavenumber ranges 1 to 28, where VIP is 1.00 or greater, as listed in Table 3. The method according to any one of claims 1 to 4.

13. The first set of wavenumber ranges of the Raman spectrum measured to obtain the first wavenumber intensity dataset of the sample includes five or more wavenumber ranges from 29 to 59, as listed in Table 3, where the Variable Importance Index (VIP) is 1.25 or greater, or The multiple wavenumber ranges of the measured Raman spectrum include 10 or more wavenumber ranges from 29 to 59, where VIP is 1.25 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum are as listed in Table 3, VI P is 1.25 or greater, and it contains 15 or more elements within the wavenumber range of 29 to 59, or The multiple wavenumber ranges of the measured Raman spectrum include 20 or more wavenumber ranges from 29 to 59 where VIP is 1.25 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 25 or more wavenumber ranges from 29 to 59, where VIP is 1.25 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 30 of the wavenumber ranges 29 to 59 where VIP is 1.25 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include all 31 of the wavenumber ranges 29 to 59, where VIP is 1.25 or greater, as listed in Table 3. The method according to any one of claims 1 to 4.

14. The first set of wavenumber ranges of the Raman spectrum measured to obtain the first wavenumber intensity dataset of the sample includes five or more wavenumber ranges from 60 to 81, with a Variable Importance Index (VIP) of 1.50 or higher, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 10 or more wavenumber ranges from 60 to 81 where VIP is 1.50 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 15 or more wavenumber ranges from 60 to 81 where VIP is 1.50 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include 20 or more wavenumber ranges from 60 to 81 where VIP is 1.50 or greater, as listed in Table 3, or The multiple wavenumber ranges of the measured Raman spectrum include all 22 of the wavenumber ranges 60 to 81 in which VIP is 1.50 or greater, as listed in Table 3. The method according to any one of claims 1 to 4.

15. The method according to any one of claims 12 to 14, wherein the virus in the sample is a lentivirus.

16. The method according to any one of the preceding claims, wherein the nucleic acid comprises a viral DNA genome, or the nucleic acid comprises a viral RNA genome.

17. The method according to any one of the preceding claims, wherein the one or more viral structural molecules comprises one or more nucleoproteins and / or one or more viral proteins such as capsomeres; one or more viral glycans; one or more glycosylated viral molecules such as glycosylated viral proteins; and / or one or more viral lipids.

18. The method according to any one of the prior claims, wherein the ratio provides a measure of functional viral titer.

19. The method according to any one of the preceding claims, wherein the sample is a virus culture, and optionally the virus culture is contained in a bioreactor.

20. The method according to claim 19, wherein the steps of irradiating the virus culture with a light source and measuring the total intensity of the Raman scattered light are performed directly (in situ) on the culture medium of the virus culture.

21. The method according to claim 19, wherein the steps of irradiating the virus culture with a light source and measuring the total intensity of the Raman scattered light are performed directly (excitually) on an aliquot of culture medium taken from the virus culture.

22. The above method is a method for a virus containing one or more viral structural molecules at the initial point in time. A method comprising a first step of determining the ratio of viral nucleic acids and one or more further steps of determining the ratio of viral nucleic acids to viruses containing one or more viral structural molecules at a later time point, further comprising measuring the change in the ratio of viral nucleic acids to viruses containing one or more viral structural molecules in the sample between time points, each step being performed by the method of any one of the preceding claims, preferably each step being performed by the same method.

23. The method according to claim 22, wherein the ratio of viral nucleic acid to a virus comprising one or more viral structural molecules is repeatedly determined over a period of time to provide a measure of change in the ratio in real time.

24. The method according to claim 22 or 23, wherein the change in the ratio in the sample is used to determine the initiation stage, production stage, and / or steady-state stage of the virus production process.

25. The method according to any one of the preceding claims, used to determine the optimal conditions for a virus production process.

26. The method according to any one of the preceding claims, used for evaluating processes downstream of a virus production process.

27. The method according to any one of the preceding claims, comprising the step of comparing the ratio obtained thereby with a ratio obtained from the same sample by an alternative method, wherein the alternative method is optionally qPCR, RT-qPCR, ELISA, or visual determination by transmission electron microscopy.

28. A method for determining the degree of viral infection of an individual using Raman spectroscopy, comprising the step of determining the ratio of viral nucleic acid to a virus containing one or more viral structural molecules in a sample by performing the method according to any one of claims 1 to 23, wherein the sample is a sample previously obtained from the individual.

29. The method according to claim 28, wherein the sample is a sample of blood, saliva, sputum, plasma, serum, cerebrospinal fluid, urine, or feces.

30. The method according to claim 28 or 29, wherein the ratio in the sample from the subject is compared with one or more ratio measurements previously obtained for the infection of the individual in order to provide a prognosis for the infection stage of the individual.

31. A method for determining the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in a sample using Raman spectroscopy, (a) (i) Providing a first wavenumber intensity dataset of the sample, wherein the first dataset is obtained by irradiating the sample with a light source and measuring the total intensity of Raman scattered light in each of a first plurality of wavenumber ranges, the first plurality of wavenumber ranges in the Raman spectrum being selected as characteristic of the viral nucleic acid in the sample, (ii) The step of performing a first set of mathematical data processing steps on the first wavenumber intensity dataset, (iii) A step of determining the nucleic acid content of the sample based on the output of the first set of mathematical data processing steps, (b) (i) Providing a second wavenumber intensity dataset of the sample, the second dataset being obtained by irradiating the sample with a light source and measuring the total intensity of Raman scattered light in each of a second plurality of wavenumber ranges, wherein the second plurality of wavenumber ranges in the Raman spectrum are selected as characteristic of one or more viral structural molecules of the virus in the sample; (ii) The step of performing a second set of mathematical data processing steps on the second wavenumber intensity dataset, (iii) A step of determining the virus content in the sample, which includes the one or more viral structural molecules, based on the output of the second set of mathematical data processing steps, (c) The step of determining the ratio of viral nucleic acid to a virus containing one or more viral structural molecules in the sample, based on the values ​​determined in steps (a)(iii) and (b)(iii), method.

32. The method according to claim 31, wherein the method is carried out according to the steps defined in any one of claims 2 to 30.

33. Use of Raman spectroscopy to determine the ratio of viral nucleic acid to viruses containing one or more viral structural molecules in a sample.

34. The use according to claim 33, wherein the ratio is determined based on a measurement of the intensity of Raman scattered light obtained from the sample after irradiation of the sample with a light source, the intensity of the Raman scattered light being measured from a first plurality of wavenumber ranges of the Raman spectrum characteristic of the viral nucleic acid in the sample and from a second plurality of wavenumber ranges of the Raman spectrum characteristic of the one or more viral structural molecules of the virus in the sample.

35. The use according to claim 34, wherein the sample is a virus culture, and optionally, the virus culture is contained in a bioreactor.

36. The use according to claim 35, wherein the step of measuring the total intensity of the Raman scattered light is performed directly (in situ) on the culture medium of the virus culture.

37. The use according to claim 35, wherein the step of measuring the total intensity of the Raman scattered light is performed directly (excisively) on an aliquot of the culture medium taken from the virus culture.

38. The use according to any one of claims 34 to 37, wherein the ratio in the sample is determined at an initial point in time and at one or more subsequent points in time, and the change in the ratio in the sample between points in time is calculated.

39. The use according to claim 38, wherein the ratio in the sample is repeatedly quantified so that a measure of the change in the ratio is provided in real time.

40. The use according to any one of claims 33 to 39, wherein the ratio in the sample is determined by performing the method according to any one of claims 1 to 18.

41. The method or use according to any one of the preceding claims, wherein the virus in the sample is not HIV-1 or HIV-1 virus-like particles (HIV-1 VLPs).

42. The method or use of the preceding claim, wherein the Raman spectroscopy is not surface-enhanced Raman spectroscopy.

43. A method for constructing a multivariate data processing model that can determine the virus content in a sample, including one or more viral structural molecules, from a wavenumber intensity dataset obtained from Raman spectroscopy of the sample, (a) The step of providing the sample and irradiating the sample with a light source, (b) A step of obtaining a wavenumber intensity dataset of the sample by measuring the total intensity of Raman scattered light in each of a plurality of wavenumber ranges, wherein the plurality of wavenumber ranges are pre-selected and characteristic of the one or more viral structural molecules of the virus in the sample, (c) The step of obtaining normalized wavenumber signal intensity data is performed by preprocessing the signal intensity data using a preprocessing analysis method such as the first derivative method, the second derivative method, the standard normal variable (SNV) method, the polynomial fitting method, the multi-polynomial fitting method, the Morifier method, the piecewise polynomial fitting (PPF) method, or the adaptive iteratively reweighted penalty least squares (airPLS) method. (d) A step of obtaining model parameters by applying a multivariate regression algorithm, such as a partial least squares (PLS) regression algorithm, to the preprocessed signal intensity data, wherein the PLS algorithm is a nonlinear iterative partial least squares (NIPALS) regression algorithm or a neural network, and calibration is performed, and the preprocessed signal intensity data is compared with viral titer data obtained under the same sample conditions using non-Raman spectroscopy such as qPCR, RT-qPCR, ELISA, or by visual determination by transmission electron microscopy, (e) The step of estimating a response value using the model parameters obtained from the preprocessed data, (f) Includes the steps of variable selection, performing a variable importance index (VIP) on an optional basis, and identifying Raman spectral variables, (g) Optionally, perform one or more further rounds of modeling by reapplying steps (d) through (f), including the step of removing non-essential variables. The virus content, which includes one or more viral structural molecules in the sample, is determined using the model parameters obtained for the identified Raman spectral variables derived from the multivariate data processing model. method.

44. A method for constructing one or more multivariate data processing models that can determine the ratio of viral nucleic acids to viruses containing one or more viral structural molecules in a sample from wavenumber intensity datasets obtained from Raman spectroscopy of the sample, (a) The step of providing the sample and irradiating the sample with a light source, (b) (i) a step of obtaining a first wavenumber intensity dataset of the sample by measuring the total intensity of Raman scattered light in each of a first plurality of wavenumber ranges, wherein the first plurality of wavenumber ranges are pre-selected and characteristic of the viral nucleic acid in the sample. (ii) A step of obtaining a second wavenumber intensity dataset of the sample by measuring the total intensity of Raman scattered light in each of a second plurality of wavenumber ranges, wherein the first plurality of wavenumber ranges are pre-selected and characteristic of the one or more viral structural molecules of the virus in the sample. (c) The step of obtaining normalized wavenumber signal intensity data of the first and second wavenumber intensity datasets, which includes the first derivative method, the second derivative method, and the standard normal variable ( This is performed by preprocessing signal intensity data using preprocessing analysis methods such as the SNV method, polynomial fitting method, multi-polynomial fitting method, Molyfire method, piecewise polynomial fitting (PPF) method, or adaptive iteratively reweighted penalty least squares (airPLS) method. (d) Obtaining model parameters to be applied to the first and second wavenumber intensity datasets by applying a multivariate regression algorithm, such as a partial least squares (PLS) regression algorithm, to each of the preprocessed signal intensity datasets, wherein optionally the PLS algorithm is a nonlinear iterative partial least squares (NIPALS) regression algorithm or a neural network, calibration is performed, and the preprocessed signal intensity data is compared with viral titer data obtained under the same sample conditions using non-Raman spectroscopy such as qPCR, RT-qPCR, ELISA, or by visual determination by transmission electron microscopy. (e) The step of estimating a response value using the model parameters obtained from each of the preprocessed datasets, (f) Includes the steps of variable selection, performing a variable importance index (VIP) on an optional basis, and identifying Raman spectral variables, (g) Optionally, perform one or more further rounds of modeling on any dataset by reapplying steps (d) through (f), including the step of removing non-essential variables. The ratio of viral nucleic acid to a virus containing one or more viral structural molecules in the sample is determined using the model parameters obtained for the identified Raman spectral variables derived from the multivariate data processing model. method.

Images