Automated calibration and maintenance of Raman spectroscopy models for real-time prediction

Abstract

A method for monitoring and / or controlling a biopharmaceutical process includes identifying query points associated with a scan of the process with a spectroscopy system (e.g., a Raman spectroscopy system) and querying an observational database containing observational data sets associated with past observations of the biopharmaceutical process. Each of the observational data sets includes spectral data and corresponding actual analytical measurements. Querying the observational database includes selecting, as training data, data sets from the observational data sets that satisfy one or more relevance criteria with respect to the query points. The method also includes calibrating a local model specific to the biopharmaceutical process using the selected training data. The local model (e.g., a Gaussian process model) is trained to predict the analytical measurement based on the spectral data input. The method also includes predicting the analytical measurement of the biopharmaceutical process using the local model.

Description

[Technical field]

[0001] CROSS-REFERENCE TO RELATED APPLICATIONS Priority is claimed to U.S. Provisional Patent Application No. 62 / 749,359, filed October 23, 2018, U.S. Provisional Patent Application No. 62 / 833,044, filed April 12, 2019, and U.S. Provisional Patent Application No. 62 / 864,565, filed June 21, 2019, each of which is incorporated herein by reference in its entirety.

[0002] This application relates generally to monitoring and / or control of biopharmaceutical processes using spectroscopic techniques such as Raman spectroscopy, and more particularly to on-line calibration and maintenance of predictive models. [Background technology]

[0003] The stable production of biotherapeutic proteins through biopharmaceutical processes generally requires that bioreactors maintain balanced and constant parameters (e.g., cell metabolic concentrations), which in turn requires strict process monitoring and control. To meet these demands, process analytical engineering (PAT) tools are increasingly being adopted. Online monitoring of pH, dissolved oxygen, and cell culture temperature are some examples of traditional PAT tools that have been used in feedback control systems. In recent years, other in-process probes have been investigated and deployed for continuous monitoring of more complex species such as viable cell density (VCD), glucose, lactate, and other important cell metabolites, amino acids, titers, and critical quality attributes.

[0004] Raman spectroscopy is a common PAT tool that is widely used for online monitoring in biopharmaceutical manufacturing. It is an optical method that allows non-destructive analysis of chemical composition and molecular structure. In Raman spectroscopy, incident laser light is inelastically scattered by molecular vibrational modes. The frequency difference between the incident and scattered photons is termed the "Raman shift", and the vector of the Raman shift versus intensity level (referred to herein as the "Raman spectrum", "Raman scan", or "Raman scan vector") can be analyzed to identify the chemical composition and molecular structure of a sample. Over the past three decades, advances in laser sampling and detector technology have led to a rapid increase in the application of Raman spectroscopy in polymer, pharmaceutical, and biopharmaceutical manufacturing, and biomedical analysis. These technological advances have made Raman spectroscopy a practical analytical technique that is now used both in and outside the laboratory. Since the application of in-situ Raman measurements in biopharmaceutical manufacturing was first reported, it has been adopted to provide online, real-time predictions of several important process conditions, such as glucose, lactate, glutamate, glutamine, ammonia, and VCD. These predictions are typically based on calibration or soft-sensor models that are constructed in an offline setting based on analytical measurements from an analytical instrument. Partial least squares (PLS) and multiple linear regression modeling methods are commonly used to correlate Raman spectra with analytical measurements. These models typically require pre-processing filtering of the Raman scan before calibrating against the analytical measurements. Once the calibration model is trained, the model is implemented in a real-time setting to provide in-situ measurements for process monitoring and / or control.

[0005] Raman model calibration for biopharmaceutical applications is not straightforward, as biopharmaceutical processes typically operate under strict constraints and regulations. The current state-of-the-art approach for Raman model calibration in the biopharmaceutical industry is to first perform multiple campaign tests to generate relevant data that are used to correlate Raman spectra to analytical measurements. These tests are both costly and time-consuming, as each campaign may span, for example, 2-4 weeks in a laboratory environment. Furthermore, only limited samples may be available to the analytical instruments (e.g., to ensure that lab-scale bioreactors maintain a healthy amount of viable cells). In fact, it is not uncommon to only have one or two measurements available each day from in-line or offline analytical instruments. To further exacerbate the situation, current best practices result in calibration models that are tied to a specific process, a specific formulation or profile of the bioreactor media, and specific operating conditions. Thus, if any of the aforementioned variables change, the model needs to be recalibrated based on the new data. In fact, both Raman model calibration and model maintenance require significant resource allocation and are typically performed in an offline setting. Approaches have been proposed to adapt the model to new operating conditions (e.g., recursive, moving window, and time-lag methods), but these methods may not be able to adequately handle rapid process changes.

[0006] There are many publications that describe general Raman models based on traditional chemometric methods (e.g., PLS modeling) for multiple molecules. However, these general models assume that the processes use similar, if not identical, media formulations and / or running process conditions. The media and processes are usually platformed with little or no variation. The drawback of this kind of general model is that the general model loses precision and accuracy once the process deviates from the standard or the training data set includes too broad a process range to account for the variation between different molecules (e.g., media additives, process duration, and / or other process changes). Thus, these "general" models are only general within the strict boundaries described. See Mehdizaheh et al., Biotechnolo.Prog.31(4):1004-1013, 2015; Webster et al., Biotechnol.Prog.34(3):730-737, 2018. [Prior art documents] [Non-patent literature]

[0007] [Non-Patent Document 1] Mehdizaheh et al., Biotechnolo.Prog.31(4):1004-1013, 2015 [Non-Patent Document 2] Webster et al., Biotechnol.Prog.34(3):730-737, 2018 Summary of the Invention [Means for solving the problem]

[0008] The term "biopharmaceutical process" refers to processes used in biopharmaceutical manufacturing, such as cell culture processes that produce a desired recombinant protein. Cell culture occurs in a cell culture vessel, such as a bioreactor, under conditions that support the growth and maintenance of an organism engineered to express the protein. During recombinant protein production, the concentrations of media components, including nutrients and metabolites (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+ and other nutrients or metabolites), media conditions (pH, pCO2, etc.), and the like are all controlled. 2 , pO 2 Process parameters such as temperature, osmolarity, etc., and cell and / or protein parameters (e.g., viable cell density (VCD), titer, cell condition, critical quality attributes, etc.) are monitored to control and / or maintain the cell culture process.

[0009] To address some of the above-mentioned limitations of current best industrial practices, embodiments described herein relate to systems and methods that improve upon conventional techniques for spectroscopic analysis of biopharmaceutical processes, such as Raman spectroscopy. In particular, a "Just-in-Time Learning" (JITL) platform is used to build and maintain calibration models (e.g., Raman calibration models) in real time for biopharmaceutical applications. JITL is a nonlinear modeling platform based on local modeling and database sampling techniques. Unlike other machine learning methods, JITL generally assumes that all available observational data is stored in a central database and models are dynamically built in real time based on queries using the most relevant data from the database. This allows for good approximation of complex process dynamics with relatively simple local models. Under the JITL framework, libraries may contain not only spectral data of a single process operating under specific operating conditions, but also data for different processes, different media profiles, and / or different operating conditions. This can significantly reduce the time required to calibrate and maintain models, especially for pipeline pharmaceutical products that may have little or no past manufacturing history.

[0010] JITL's platform maintains a dynamic library that may be updated whenever new analytical measurements become available. Furthermore, to ensure that the local model is adapted to new process conditions, the last available analytical measurement (e.g., for the currently monitored product) may always be included in the training set for local modeling. This allows the local model to adapt more quickly to new conditions, or new product lines with no history. Using this approach, both model calibration and model maintenance may be automated, potentially significantly reducing the time and costs (e.g., material and labor costs) associated with routine calibration in conventional systems. Furthermore, the ability to provide confidence limits (or other confidence indicators, such as confidence scores) for model predictions may enable robust monitoring and control strategies.

[0011] In some embodiments, Gaussian process models are used for local modeling within the JITL framework. Gaussian process models are powerful statistical machine learning models that can efficiently capture complex nonlinear process dynamics and can easily adapt to virtually any process changes. In contrast to PLS, Principal Component Regression (PCR), and other types of regression models, Gaussian process models are nonparametric methods and are much more capable of capturing complex correlations between Raman spectra and analytical measurements from limited data sets. Furthermore, Gaussian process models generally do not require pre-processing filtering of Raman scans. Thus, in some embodiments, Gaussian process models are instead calibrated on raw Raman scans (on a logarithmic scale), which may save many steps in the model calibration / maintenance process. Furthermore, Gaussian process models provide prediction-centric confidence limits that may be extremely difficult to obtain using PLS or PCR models. Confidence limits may be particularly useful for designing optimal sampling strategies for analytical instruments and / or for implementing closed-loop control (e.g., model predictive control, or MPC), e.g., to avoid changes made based on unreliable predictions.

[0012] Although JITL is a nonlinear modeling framework, and the approach described above provides some adaptability by updating a dynamic library with recent analytical measurements, JITL alone may not be adaptable enough to account for time-varying process conditions (e.g., rapid changes to setpoint or other process conditions). In particular, local models calibrated with JITL may not be able to take advantage of recent samples. For example, recent samples may not be able to meet similarity criteria based purely on "spatial" similarity (e.g., similarity of Raman scans), especially if there have been recent rapid changes in process conditions. Modified JITL techniques are also described herein that can better exploit information provided by recent samples (regardless of spatial similarity) and thus better adapt to time-varying process changes. In particular, "adaptive" JITL (A-JITL) and "spatio-temporal" JITL (ST-JITL) techniques for model calibration and maintenance are described herein.

[0013] Real-time model maintenance, where local models can learn from the latest analytical measurements and thereby rapidly adapt to time-varying conditions, can be critical to the success of JITL techniques. However, frequent access to analytical instruments / measurements (e.g., analyzing offline samples) tends to be extremely resource intensive. To minimize such resource usage without unduly degrading model performance, a performance-based model maintenance protocol may be implemented in which the system schedules / trigger analytical measurements in response to identifying current model performance as unacceptable / unreliable.

[0014] Those skilled in the art will appreciate that the figures described herein are included for illustrative purposes and are not intended to limit the disclosure. The figures are not necessarily to scale, with emphasis instead being placed upon illustrating the principles of the disclosure. It should be understood that in some instances, various aspects of the described implementations may be shown exaggerated or enlarged to facilitate understanding of the described implementations. In the drawings, like reference numbers generally refer to functionally similar and / or structurally similar components throughout the various views. [Brief description of the drawings]

[0015] [Figure 1] FIG. 1 is a simplified block diagram of an example Raman spectroscopy system that may be used to predict analytical measurements of a biopharmaceutical process. [Diagram 2] FIG. 1 is a simplified block diagram of an example Raman spectroscopy system that may be used to predict analytical measurements of a biopharmaceutical process for closed-loop control of glucose concentration. [Diagram 3] 13 shows experimental results for closed-loop control of glucose concentration using an example implementation of the Raman spectroscopy system described herein. [Figure 4] 1 illustrates an example data flow that may occur when analyzing a biopharmaceutical process using just-in-time learning (JITL) techniques. [Diagram 5] 1 illustrates an example data flow that may occur when analyzing a biopharmaceutical process using adaptive JITL (A-JITL) techniques. [Figure 6] 1 illustrates an example data flow that may occur when analyzing a biopharmaceutical process using spatio-temporal JITL (ST-JITL) techniques. [Figure 7] FIG. 1 is a flow diagram of an example method for analyzing a biopharmaceutical process. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0016] The various concepts introduced above and further detailed below can be implemented in any of numerous ways, and the concepts described are not limited to any particular manner of implementation. For illustrative purposes, example implementations are provided.

[0017] Figure 1 is a simplified block diagram of an example Raman spectroscopy system 100 that may be used to predict analytical measurements of a biopharmaceutical process. Although Figure 1 illustrates system 100 performing Raman spectroscopy, it will be appreciated that in other embodiments, system 100 may perform other spectroscopy techniques suitable for analyzing biopharmaceutical processes, such as, for example, near-infrared (NIR) spectroscopy.

[0018] The system 100 includes a bioreactor 102, one or more analytical instruments 104, a Raman analyzer 106 having a Raman probe 108, a computer 110, and a database server 112 coupled to the computer 110 via a network 114. The bioreactor 102 may be any suitable vessel, device, or system that supports a biologically active environment that may include living organisms in the medium and / or substances derived therefrom (e.g., cell cultures). The bioreactor 102 may include recombinant proteins expressed by the cell cultures, such as for research purposes, clinical use, commercial sale, or other distribution, etc. Depending on the biopharmaceutical process being monitored, the medium may include a particular fluid (e.g., "broth") and particular nutrients, and may have target medium state parameters, such as a target pH level or range, a target temperature or temperature range, etc. The medium may also include organisms and substances derived from the organisms, such as metabolites and recombinant proteins. Collectively, the contents and parameters / characteristics of the medium are referred to herein as a "media profile."

[0019] The analytical instrument 104 may be any in-line, at-line, and / or offline instrument or instruments configured to measure one or more characteristics or parameters of the biologically active contents within the bioreactor 102 based on samples taken therefrom. For example, the analytical instrument 104 may measure nutrient and / or metabolite levels (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+, etc.) and media condition parameters (pH, pCO2, 2 , pO 2 The analytical instruments 104 may measure one or more media component concentrations such as osmolality, temperature, osmolality, etc. Additionally or alternatively, the analytical instruments 104 may measure osmolality, viable cell density (VCD), titer, key quality attributes, cell state (e.g., cell cycle), and / or other characteristics or parameters associated with the contents of the bioreactor 102. As a more detailed example, a sample may be taken, centrifuged, purified through multiple columns, passed through a first of the analytical instruments 104 (e.g., a high performance liquid chromatography (HPLC) or ultra performance liquid chromatography (UPLC) instrument) and subsequently through a second of the analytical instruments 104 (e.g., a mass spectrometer), with the first and second analytical instruments 104 providing analytical measurements. One, some, or all of the analytical instruments 104 may use destructive analytical techniques.

[0020] The Raman analyzer 106 may include a spectrograph device coupled to the Raman probe 108 (or, in some implementations, multiple Raman probes). The Raman analyzer 106 may include a laser source that transmits laser light to the Raman probe 108 via a fiber optic cable, and may also include, for example, a charge-coupled device (CCD) or other suitable camera / recording device that records signals received from the Raman probe 108 via another channel of the fiber optic cable. Alternatively, the laser source may be integrated within the Raman probe 108 itself. The Raman probe 108 may be an immersion probe, or any other suitable type of probe (e.g., a reflection probe and a transmission probe).

[0021] Collectively, the Raman analyzer 106 and Raman probe 108 are configured to non-destructively scan the biologically active content during a biopharmaceutical process in the bioreactor 102 by exciting, observing, and recording a molecular "fingerprint" of the biopharmaceutical process. The molecular fingerprint corresponds to vibrational, rotational, and / or other low frequency modes of molecules within the biologically active content within the biopharmaceutical process when the bioreactor contents are excited by laser light delivered by the Raman probe 108. As a result of this scanning process, the Raman analyzer 106 generates one or more Raman scan vectors, each of which represents intensity as a function of Raman shift (frequency).

[0022] The computer 110 is coupled to the Raman analyzer 106 and the analytical instrument 104 and is generally configured to analyze the Raman scan vectors generated by the Raman analyzer 106 to predict one or more analytical measurements of the biopharmaceutical process. For example, the computer 110 may analyze the Raman scan vectors to predict the same types of analytical measurements made by the analytical instrument 104. As a more detailed example, the computer 110 may predict a glucose concentration while the analytical instrument 104 actually measures the glucose concentration. However, the analytical instrument 104 may perform relatively infrequent "offline" analytical measurements of samples extracted from the bioreactor 102 (e.g., due to limited amounts of media from the biopharmaceutical process and / or due to higher costs of making such measurements), whereas the computer 110 may perform relatively frequent "online" predictions of analytical measurements in real time. The computer 110 may also be configured to transmit the analytical measurements made by the analytical instrument 104 to the database server 112 via the network 114, as described in more detail below.

[0023] In the exemplary embodiment shown in FIG. 1, computer 110 includes a processing unit 120, a network interface 122, a display 124, a user input device 126, and a memory 128. Processing unit 120 includes one or more processors, each of which may be a programmable microprocessor that executes software instructions stored in memory 128 to perform some or all of the functions of computer 110 as described herein. Alternatively, one, some, or all of the processors in processing unit 120 may be other types of processors (e.g., application specific integrated circuits (ASICs), field programmable gate arrays (FPGAs), etc.), and the functions of computer 110 described herein may instead be implemented partially or entirely in hardware. Memory 128 may include one or more physical memory devices or units, including volatile and / or non-volatile memory. Any suitable type or types of memory may be used, such as read only memory (ROM), solid state drive (SSD), hard disk drive (HDD), etc.

[0024] Network interface 122 may include any suitable hardware (e.g., front-end transmitter and receiver hardware), firmware, and / or software configured to communicate over network 114 using one or more communications protocols. For example, network interface 122 may be or may include an Ethernet interface. Network 114 may be a single communications network or may include multiple communications networks of one or more types (e.g., one or more wired and / or wireless local area networks (LANs) and / or one or more wired and / or wireless wide area networks (WANs), such as the Internet).

[0025] The display 124 may use any suitable display technology (e.g., LED, OLED, LCD, etc.) to present information to a user, and the user input device 126 may be a keyboard or other suitable input device. In some embodiments, the display 124 and the user input device 126 are integrated into a single device (e.g., a touch screen display). In general, the display 124 and the user input device 126 may be combined to allow a user to interact with a graphical user interface (GUI) provided by the computer 110, for purposes such as, for example, to manually monitor various processes running within the system 100. However, in some embodiments, the computer 110 does not include the display 124 and / or the user input device 126, or one or both of the display 124 and the user input device 126 are included in another computer or system that is communicatively coupled to the computer 110 (e.g., in some embodiments where predictions are sent directly to a control system implementing closed-loop control).

[0026] The memory 128 stores instructions for one or more software applications, including a just-in-time learning (JITL) prediction application 130. The JITL prediction application 130, when executed by the processing unit 120, is generally configured to predict analytical measurements of a biopharmaceutical process in the bioreactor 102 by calibrating a local model 132 and by analyzing Raman scan vectors generated by the Raman analyzer 106 with the local model 132. Depending on how frequently the Raman analyzer 106 generates such scan vectors, the JITL prediction application 130 may predict analytical measurements periodically or on another suitable time basis. The Raman analyzer 106 may itself control when the scan vectors are generated, or the computer 110 may trigger the generation of the scan vectors by sending a command to the Raman analyzer 106. The JITL prediction application 130 may predict only a single type of analytical measurement based on each scan vector (e.g., only glucose concentration), or may predict multiple types of analytical measurements based on each scan vector (e.g., glucose concentration and viable cell density). In another embodiment, multiple different JITL prediction applications (e.g., each similar to JITL prediction application 130) each generate a different local model that predicts a different type of analytical measurement, all based on the same scan vector. JITL prediction applications 130 and local models 132 are described in more detail below.

[0027] The database server 112 may be remote from the computer 110 (e.g., such that a local facility may include only the bioreactor 102, the analytical instrument 104, the Raman analyzer 106 with the Raman probe 108, and the computer 110) and may include or be communicatively coupled to an observational database 136 that stores observational data sets related to past observations, as seen in FIG. 1. Each observational data set in the observational database 136 may include spectral data (e.g., one or more Raman scan vectors of the type generated by the Raman analyzer 106) and one or more corresponding analytical measurements (e.g., one or more measurements of the type generated by the analytical instrument 104). In some embodiments and / or scenarios, past observations may have been collected for a number of different biopharmaceutical processes, under a number of different operating conditions (e.g., different metabolite concentration set points) and / or with a number of different media profiles (e.g., different fluids, nutrients, pH levels, temperatures, etc.). In general, it may be desirable for the observational database 136 to represent a widely diverse array of processes, operating conditions, and media profiles. The observational database 136 may or may not store information indicative of these processes, cell lines, proteins, metabolites, operating conditions, and / or media profiles, depending on the embodiment (as discussed further below). In some embodiments, the database server 112 is remotely coupled to multiple other computers similar to the computer 110 via the network 114 and / or other networks. This may be desirable to collect multiple observational data sets for storage in the observational database 136. However, in other embodiments, the system 100 does not include a database server 112, and the computer 110 directly accesses the local observational database 136.

[0028] It will be appreciated that other configurations and / or components may be used in place of those shown in Figure 1. For example, a different computer (not shown in Figure 1) may transmit measurements provided by analytical instruments 104 to database server 112, one or more additional computing devices or systems may act as intermediaries between computer 110 and database server 112, some or all of the functionality of computer 110 as described herein may instead be performed remotely by database server 112 and / or another remote server, etc.

[0029] During run-time operation of the system 100, the Raman analyzer 106 and the Raman probe 108 are used to scan (i.e., generate Raman scan vectors for) the biopharmaceutical process in the bioreactor 102, which are then transmitted from the Raman analyzer 106 to the computer 110. The Raman analyzer 106 and the Raman probe 108 may provide scan vectors to support predictions (made by the JITL prediction application 130) according to a pre-defined schedule of monitoring periods, such as once per minute or once per hour. Alternatively, predictions may be made at irregular intervals (e.g., in response to certain process-based triggers, such as changes in measured pH levels and / or temperature), such that each monitoring period has a variable or uncertain duration. In some embodiments, the Raman analyzer 106 may transmit only one scan vector per monitoring period to the computer 110, or multiple scan vectors per monitoring period to the computer 110, depending on how many scan vectors the local model 132 accepts as input for a single prediction. Multiple scan vectors, for example, may improve the predictive accuracy of the local model 132.

[0030] A query unit 140 of the JITL prediction application 130 generates query points that are used to query the observation database 136 using the scan vectors received during a single monitoring period. In some embodiments, the query points (i.e., the data defining the query points) include only data representative of the Raman scan vectors received from the Raman analyzer 106 (e.g., the intensity / frequency tuples that make up each scan vector). In other embodiments, the query points also include one or more other types of information. For example, the query points may also include data representative of operating conditions associated with the process (e.g., metabolite concentration set points in a control system, or laser light wavelengths and / or intensities associated with the Raman analyzer 106 or Raman probe 108, etc.), data representative of a media profile for a biopharmaceutical process media (e.g., fluid type, nutrient type or concentration, pH level, etc.), and / or other data (e.g., indicators of cell lines, proteins, or metabolites associated with the biopharmaceutical process).

[0031] In general, the query points may include data representing the same vectors, parameters, and / or classifications that the local model 132 uses as inputs (i.e., as the feature set of the local model 132). Using multiple different data types for the feature set may improve the accuracy of the analytical measurement predictions made by the local model 132. However, because each observational data set in the observational database 136 should generally include the same vectors, parameters, and / or classifications as the feature set, it may be preferable to limit the query points, and the feature set / inputs of the local model 132, to include only one or more Raman scan vectors. This may provide various advantages, such as allowing for the collection of more information for storage in the observational database 136 and / or simplifying the collection of that information. For example, if only Raman scan vectors are used, an observational data set may be included in the observational database 136 even if little or nothing is known about the processes, cell lines, proteins, metabolites, operating conditions, and / or media profiles that existed when the data set was collected.

[0032] The query unit 140 then queries the observation database 136 using the generated query points. In the exemplary embodiment of FIG. 1, the query unit 140 accomplishes this by having the network interface 122 transmit the query points (e.g., in a query message) over the network 114 to the database server 112, which in turn causes the database server 112 to retrieve the appropriate data from the observation database 136. However, in embodiments in which the observation database 136 is instead contained in (or contained in a memory communicatively coupled to) the computer 110, the query unit 140 may instead query the observation database 136 more directly. For ease of explanation, the remaining description of FIG. 1 will assume that the observation database 136 is coupled to the database server 112 as shown in FIG. 1. However, one skilled in the art will readily appreciate how the communication paths would be different if the observation database 136 was local to the computer 110 or in another suitable location within the system architecture.

[0033] After receiving the query points, the database server 112 uses the query points to select relevant observational data sets from the observational database 136 that are useful as training data for the local model 132. Depending on the embodiment, the database server 112 may apply any suitable relevance criteria to identify which observational data sets are "relevant". In one embodiment, for example, the query points include a single Raman scan vector, and the database server 112 identifies whether a given observational data set is relevant by calculating the Euclidean distance between the Raman scan vector of that observational data set and the Raman scan vector of the query point. If the Euclidean distance is below some predefined threshold (or below a variable threshold, such as a threshold calculated based on the average Euclidean distance between the query point scan vector and all observational data set scan vectors), the observational data set is identified as a relevant observational data set. Those skilled in the art will understand how such an approach can be easily extended to embodiments in which the query points (and each observational data set) include multiple Raman scan vectors. In some circumstances, using Euclidean distance to select relevant observational data sets may be a suboptimal technique. However, when the local model 132 is a Gaussian process model (as discussed below), using Euclidean distance as the relevance criterion may be particularly advantageous because Gaussian process models with radial basis functions or squared exponential kernels are themselves based on Euclidean distance. Nevertheless, in other embodiments, other relevance criteria (e.g., angle-based or correlation-based criteria, etc.) may be applied. Of course, in embodiments where the local model 132 also accepts other information as input / feature sets (e.g., operating conditions, media profiles, process data, cell line information, protein information, and / or metabolite information, etc.), more sophisticated techniques may be used to identify "relevant" observational data sets.In some embodiments, database server 112 selects only a predetermined number of relevant observational data sets in response to a single query, or selects no more than some maximum allowable number of relevant observational data sets, ensuring that only a relatively small subset of all data sets in observational database 136 is searched, however, in other embodiments, database server 112 may select any number of relevant observational data sets, so long as the relevance criteria are met for each such data set.

[0034] In some embodiments, as described in more detail below (e.g., with reference to Figures 5 and 6), relevant observational data sets are selected based not only on their relevance to the query point in a "spatial" sense (e.g., similarity of Raman scan vectors), but also on their relevance in a temporal sense (e.g., which data set is the most recent, regardless of spatial similarity). These techniques may better exploit the fact that more recent analytical measurements can provide useful information, even if those recent measurements correspond to different set points, etc.

[0035] After identifying the relevant observational data sets (each of which may or may not correspond to the same process conditions as the biopharmaceutical process in the bioreactor 102 currently being monitored), the database server 112 retrieves those data sets (e.g., Raman scan vectors and corresponding analytical measurements) and transmits the retrieved data sets to the computer 110 via the network 114. The query unit 140 may then pass the relevant data sets to the local model generator 142, which uses the relevant data sets as training data to calibrate the local model 132. That is, the local model generator 142 uses the Raman scan vectors (and possibly other data) associated with each observational data set as a feature set and the analytical measurements associated with the same observational data set as labels for that feature set.

[0036] In some embodiments, as described above, the local model generator 142 builds a Gaussian process model to efficiently capture complex nonlinear process dynamics and easily adapt to virtually any process change. Unlike PLS and PCR models, the Gaussian process model uses nonparametric methods and is much more likely to capture complex nonlinear correlations between Raman scan vectors and analytical measurements, even with a very limited number of training samples. This may be particularly important in scenarios where a new product or process corresponds to only a limited number of data sets in the observation database 136. In such scenarios, the Gaussian process model can generally extract the most information from those limited data sets, along with other relevant data sets that the database server 112 selects from the observation database 136. However, in other embodiments, the local model generator 142 may instead build any other suitable type of machine learning model (e.g., recurrent neural network, convolutional neural network, etc.), so long as the training time does not exceed the minimum desired duration of the monitoring period. The local model generator 142 may construct the local model 132 such that the local model 132 can output confidence limits or other suitable indicators of prediction confidence (e.g., confidence scores). At least in comparison to PLS and PCR models, Gaussian process models are particularly suitable for providing confidence limits around analytical measurement predictions. Although various advantages of Gaussian process models over PLS and PCR models have been described, it will be appreciated that in some embodiments the local model generator 142 may construct the local model 132 using PLS or PCR modeling methods.

[0037] The local model generator 142 may build the local model 132 in an online, real-time manner such that the prediction unit 144 may then use the trained local model 132 to predict one or more analytical measurements of the biopharmaceutical process by processing the same Raman scan vectors that the query unit 140 used to generate the query points. Indeed, in some embodiments, the query unit 140 may perform a new query, and the local model generator 142 may generate a new version of the local model 132, each time the Raman analyzer 106 provides the computer 110 with a new Raman scan vector (or a new set of Raman scan vectors). However, in other embodiments, the query unit 140 performs a new query (and the local model generator 142 generates a new version of the local model 132) on an infrequent basis, such as once every 10 prediction / monitoring periods, or once every 100 prediction / monitoring periods, etc.

[0038] The database maintenance unit 146 may also cause the analytical instrument 104 to periodically collect one or more actual analytical measurements at a frequency significantly less than the monitoring period of the Raman analyzer 106 (e.g., once or twice a day, etc.). The measurements by the analytical instrument 104 may be destructive in some embodiments and may require permanent removal of the sample from the process in the bioreactor 102. At or about the time the database maintenance unit 146 causes the analytical instrument 104 to collect and provide an actual analytical measurement, the database maintenance unit 146 may also cause the Raman analyzer 106 to provide one or more Raman scan vectors. The database maintenance unit 146 may then cause the network interface 122 to transmit the Raman scan vectors and the corresponding actual analytical measurements over the network 114 to the database server 112 for storage as a new observational data set in the observational database 136. The observational database 132 may be updated according to any suitable timing, which may vary depending on the embodiment. In cases where an analytical instrument 104 outputs an actual analytical measurement within seconds of measuring a sample, for example, the observational database 132 may be updated with a new measurement almost immediately after the sample is taken. However, in certain other embodiments, the actual analytical measurement may be the result of minutes, hours, or days of processing by one or more analytical instruments 104, in which case the observational database 132 is not updated until such processing is complete. In still other embodiments, new observational data sets may be added to the observational database 132 incrementally as different ones of the analytical instruments 104 complete their respective measurements.

[0039] Thus, the observation database 136 provides a "dynamic library" of past observations that the local model generator 142 may draw from for model training. In some embodiments, the latest analytical measurements are always added to the observation database 136, and the local model generator 142 may always use the most recent observation data set in the observation database 136 when calibrating the local model 132. This may enable the local model 132 to encode process information from the recent past and rapidly adapt to new conditions or to new process conditions for which it has no history. Moreover, both the calibration and maintenance of the local model 132 may be automated. In some embodiments, the adaptability of the local model 132 is further enhanced, for example, as discussed below in connection with A-JITL and ST-JITL techniques.

[0040] In some embodiments, the database maintenance unit 146 may cause the analytical instrument 104 to collect and provide actual analytical measurements on some other time basis or condition, such as current model performance. For example, if the local model 132 outputs a confidence interval (e.g., a range of values ​​around the predicted value that has a 95% probability or confidence that the actual / measured value is included) or some other confidence indicator with the prediction (e.g., if the local model 132 is a Gaussian process model), and if the confidence indicator reveals a particularly unreliable prediction (e.g., if the interval / range exceeds a threshold width / range, etc.), the database maintenance unit 146 may trigger the collection of one or more actual analytical measurements. As a more detailed example, the database maintenance unit 146 may trigger the collection of analytical measurements in response to identifying that a 95% confidence interval exceeds a predefined threshold. Optimal scheduling of analytical measurements is discussed in more detail below. After the measurements have been made, the database maintenance unit 146 may cause the network interface 122 to provide the actual analytical measurements and the corresponding Raman scan vectors to the database server 112 for the Raman analyzer 106 to generate one or more Raman scan vectors and store them as a new observational data set in the observational database 132 (e.g., in the manner discussed above). The local model generator 142 may then utilize that most recent observational data set, where appropriate (e.g., depending on its relevance to the current query or whether the embodiment always utilizes the most recent observational data set), in calibrating the local model 132.

[0041] Some or all of the processes described above may be repeated many times over the life of a biopharmaceutical process in a bioreactor to continuously monitor the process using a local model that is both fully automated and real-time in calibration and maintenance. Analytical measurements may be predicted for various purposes, depending on the embodiment and / or scenario. For example, certain parameters may be monitored (i.e., predicted) as part of a quality control process to ensure that the process is still compliant with relevant regulations. As another example, one or more parameters may be monitored / predicted to provide feedback in a closed-loop control system. For example, FIG. 2 illustrates a system 150 similar to system 100, but that attempts to control glucose concentration in a biopharmaceutical process (i.e., attempts to match a predicted glucose concentration to a desired set point within some accepted tolerance range). It will be appreciated that in other embodiments, system 150 may instead (or similarly) be used to control process parameters other than glucose level, or to control glucose levels based on predictions of one or more other process parameters (e.g., lactate level). In FIG. 2, the same reference numbers are used to indicate corresponding components of FIG. 1. For example, the JITL prediction application 130 of FIG. 2 may be the same as the JITL prediction application 130 of FIG. 1 (for clarity, various units of the JITL prediction application 130 are not shown in FIG. 2).

[0042] 2, in the system 150, the memory 128 also stores a control unit 152. The control unit 152 is configured to control the glucose pump 154, i.e., to cause the glucose pump 154 ​​to selectively introduce additional glucose into the biopharmaceutical process in the bioreactor 102. The control unit 152 may comprise, for example, software instructions executed by the processing unit 120 and / or suitable firmware and / or hardware. In some embodiments, the control unit 152 implements a model predictive control (MPC) technique that uses glucose concentration as an input in a closed-loop architecture. In embodiments in which the local model 132 provides a confidence limit or other confidence metric with each prediction (e.g., certain embodiments in which the local model 132 is a Gaussian process model), the control unit 152 may also accept the confidence metric as an input. For example, the control unit 152 may only generate control instructions for the glucose pump 154 ​​based on glucose concentration predictions that have a sufficiently high confidence index (e.g., only based on predictions associated with confidence limits that do not exceed a certain percentage or absolute measurement range, or only based on predictions associated with confidence scores above a certain minimum threshold score), or may increase and / or decrease the weighting of certain predictions based on their confidence index.

[0043] Figure 3 shows experimental results 200 for one embodiment in which the JITL technique was used to calibrate and maintain a local Gaussian process model. In the plot of Figure 3, the horizontal dashed line 202 represents the glucose concentration set point, the circle 204 represents an actual measurement of glucose concentration (e.g., made by an analytical instrument similar to one of the analytical instruments 104 of Figure 1), the solid line 206 represents a predicted measurement of glucose concentration (e.g., as predicted by a model similar to the local model 132), and the shaded area 208 represents the confidence limits (for 95% confidence) associated with the predicted measurements. As can be seen in Figure 3, for a glucose concentration set point of 3 grams per liter (g / L), the predictions made using the JITL technique generally closely match the analytical measurements.

[0044] The process of executing queries and building / calibrating the local model 132 is now described in more mathematical detail with reference to one particular JITL embodiment in which the local model 132 is a Gaussian process model that uses a single Raman scan vector as input and predicts a single analytical measurement.

[0045]

number

number

number

number

number

number

number

number

number

number

number

number

[0046] Also,

number

number

number

[0047] The spectroscopic model calibration problem is then reduced using D to learn the following latent Gaussian process function:

number

number

number

number

number

[0048] For the choice of Gaussian kernel, equation (4) is a positive definite symmetric matrix such that

number

number

number

number

[0049] Given D, we aim to learn the hyperparameters of a Gaussian process that includes any other unknown model parameters. For the Gaussian process in equation (1), the set of unknown parameters is

number

number

number

number

number

number

number

[0050] Now, given equation (7),

number

number

number

number

number

number

[0051] Once the Gaussian process spectroscopic calibration model in Equation (1) is trained, it can be deployed for real-time predictive applications. As mentioned above, let D be the training data set used to train the Gaussian process model,

number

number

number

number

number

number

number

number

number

number

number

[0052] Given equation (12), the predictive posterior distribution for output b* can be calculated as follows:

number

number

number

number

number

number

number

[0053] Turning now to the selection of relevant samples (here, observational datasets) in response to a query, the problem is

number

number

number

number

number

number

[0054] An example algorithm that formally outlines the method of creating a local training set D from L, training a Gaussian process model using that training set, and making predictions using the trained model is provided in Algorithm 1 below.

Number

[0055] Turning now to FIG. 4, an exemplary data flow 250 is shown that may occur when analyzing a biopharmaceutical process using the JITL technique as described herein. The data flow 250 may occur, for example, in the system 100 of FIG. 1 or the system 150 of FIG. 2. In the data flow 250, spectral data 252 is provided by a spectrometer / probe. For example, the spectral data 252 may include Raman scan vectors generated by the Raman analyzer 106, or NIR scan vectors, etc. A query point 254 is generated (e.g., by the query unit 140) based on the spectral data 252 and is used to query a global dataset 256, which may include, for example, all observational datasets in the observational database 136. Based on the query, a local dataset 258 is identified in the global dataset 256. The local dataset 258 may be selected, for example, based on a relevance criterion (e.g., Euclidean distance), as described above.

[0056] The local data set 258 is then used as training data (e.g., by the local model generator 142) to calibrate a local model 260 (e.g., the local model 132). The local model 132 then calculates the media component concentrations, media conditions (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+ and other nutrients or metabolites, pH, pCO 2 , pO 2 , temperature, osmolality, etc.), viable cell density, titer, important quality attributes, cell condition, etc., and possibly also output confidence limits or another suitable confidence indicator.

[0057] While JITL-based local models (e.g., as in Algorithm 1 and dataflow 250) provide a robust nonlinear modeling framework, such approaches have no inherent mechanism for adapting to time-varying process changes. To address this shortcoming, some embodiments may employ an “adaptive” JITL (A-JITL) strategy. As mentioned above, new samples may be included in L as they become available. In such embodiments (i.e., when L is dynamic), L may be a function of L t In one such embodiment, a moving time window method is implemented, in which a newly acquired sample is t The oldest sample was added to L t Discarding the oldest samples is an adaptive strategy that t This may be beneficial since maintaining the size of,may be important to ensure the computational tractability of the overall,JITL framework.,However, one major concern with this approach is that simply discarding old samples,may lead to information loss, as old samples may contain,relevant information.

[0058] To avoid such information loss, in one embodiment, new samples are added to the L without removing old / existing samples. t Therefore, the central database L t will grow with an increasing number of samples as new analytical measurements become available. In cell culture process applications, a growing database may not pose any significant computational problems due to the fact that such processes are typically operated as batch processes with batch times of 2-3 weeks. This is, of course, t In addition, only a limited number of analytical measurements are typically sampled during the course of a cell culture process batch (unlike, for example, the chemical industry, where analytical measurements are sampled frequently). Thus, typically, the database L tThe small increase in size of ,does not have any significant impact on the computational stability of the overall JITL framework.

[0059] L t Although the inclusion of new samples in is important for the continuous adaptation of Algorithm 1 (above), the success of this approach relies on the selection of these new samples in the local database D for local model calibration. Algorithm 1 (e.g., line 6 of Algorithm 1), which selects samples for D from L based on Euclidean distance, can be referred to as a "relevant in space" approach because it prioritizes only samples that are related (close) in space. If new samples are not close to the query samples, such as when a sudden set point change (or other sudden process condition change) occurs, Algorithm 1 may fail to include them in D. On the other hand, recursive methods (e.g., Regularized Partial Least Squares (RPLS), Recursive Least Squares (RLS), and Recursive N-way Partial Least Squares (RNPLS)) are "relevant in time" because they prioritize the most recent measurements regardless of their relevance in space. Updating the local model with the most recent samples can enable the recursive methods to successfully adapt to the current process conditions.

[0060] One such embodiment, referred to herein as "adaptive" JITL (A-JITL), prioritizes samples that are related in both space and time.

number

number

number

number

number

number

number

[0061] L t Given and K, the objective is to select D. As mentioned above, for A-JITL, both spatial and temporal related samples are included in D. Assume that D can be decomposed as follows: D≡D S ∪D T , Equation (18) Here, D S and D. T are spatial and temporal related sets, respectively, and the goal is D S and D. T First, D S ∩DT = 0 and D contains only unique samples. S To design the Dk samples, L is determined based on a distance-based (spatial) metric such as a "similarity index" or "s-value" as follows: t is selected from. S i =sim(a i ,a * )=exp(-||a i -a * ||) Formula (19)

[0062] Equation (19) may be used, for example, as a similarity metric in the (non-adaptive) JITL technique described above. Thus, for example, the Dk samples with the largest s value are D S To include in L t may be selected from D T To design, if we assume that the last k samples from the current experiment / process are related in time, D T may be defined as equal to K in some embodiments. S Unlike the s value that identifies the membership of a sample in T Note that membership in is determined based on sampling time. Of course, in some scenarios, T Samples within the range may show large s values. Regardless of the s value, D T can only be considered to be related in time. Similarly, by construction, L t Since there is no time correlation, D S is only relevant in space. D S and D. T is defined for a given query a*, and D S The samples in were selected based on their s values ​​calculated with respect to a*, and D T Note that the samples in are selected based on their sampling times calculated relative to the sampling times of a*. For convenience, D S and D. T is generally defined as follows:

number

number

number

number

number

number

[0063] Substituting equations (20a) and (20b) into equation (18), we obtain the set D, which can be generally expressed as:

number

number

number

number

number

number

number

number

number

[0064] An example algorithm that formally outlines the A-JITL technique is provided in Algorithm 2 below.

number

[0065] Therefore, Algorithm 2 combines JITL (relevant in space) and recursive learning (relevant in time). For example, |D T For |=0, the calibration of the local model 132 using Algorithm 2 is similar to the associated JITL in space, whereas |D S If |D = 0|, then the calibration of the local model 132 using Algorithm 2 is similar to recursive learning. S | and |D T By adjusting |, you can get the right balance between (non-recursive) JITL and recursive learning.

[0066] Turning now to FIG. 5, an exemplary data flow 300 that may occur when analyzing a biopharmaceutical process using the A-JITL technique as described herein is shown. The data flow 300 may occur, for example, in the system 100 of FIG. 1 or the system 150 of FIG. 2. In the data flow 300, spectral data 302 is provided by a spectrometer / probe. For example, the spectral data 302 may include Raman scan vectors generated by the Raman analyzer 106, or NIR scan vectors, etc. Query points 304 are generated (e.g., by the query unit 140) based on the spectral data 302 and are used to query a global dataset 306 that may include, for example, all observational datasets in the observational database 136. The global dataset 306 is logically separated into the last k entries 307A (e.g., all from the current experiment / process) and all entries 307B before the last k entries 307A (e.g., from previous experiments / processes and possibly also from the current experiment / process). The value of k may be determined based on the sample count of the query point 304. As used herein, the term "sample count" may broadly refer to any indication of time or relative time associated with a given sample / observation. All entries 307A may be added to the local dataset 308 regardless of spatial similarity, while certain entries in entries 307B may be added to the local dataset 308 based on spatial similarity (e.g., Euclidean distance) to the query point 304. The local dataset 308 may be generated from entries 307A and entries 307B, for example, according to Algorithm 2.

[0067] The local data set 308 is then used as training data (e.g., by the local model generator 142) to calibrate a local model 310 (e.g., the local model 132). The local model 310 then calculates the media component concentrations, media conditions (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+ and other nutrients or metabolites, pH, pCO 2 , pO 2 , temperature, osmolality, etc.), viable cell density, titer, important quality attributes, cell condition, etc., and possibly also output confidence limits or another suitable confidence metric.

[0068] When an actual analytical measurement (e.g., a measurement made by an analytical instrument such as one of the analytical instruments 104) is available, a new entry 314 is created and added to the global dataset 306. Such measurements may be available, for example, on a periodic sampling basis (e.g., once or twice a day) and / or may become available in response to a trigger having variable timing, as discussed further below (e.g., if predictions for a certain number in a row have unacceptably wide confidence limits, etc.).

[0069] Although the inclusion of spatial and temporal related samples in D is necessary for the continuous adaptation of the A-JITL approach discussed above, the overall degree of adaptation achieved by A-JITL depends on how effectively D is utilized for local model calibration. For a query sample / point a*, the spatial related samples (a i ,b i )∈D S is the function (f(a*), f(a i )) for a query a*. i ,b i ) and the spatial correlation of (f(a*),f(a i )) are both correlated with (a i, a*). Therefore, for the selection of the Euclidean-based similarity measure in (19) and the Euclidean-based kernel in (4), S The samples at are expected to provide high functional correlation. Conversely, the time-related samples (a j ,b j )∈D T is the function (f(a*), f(a j )) may not provide a strong correlation between the D T This is because the samples in are not necessarily related in space. j ,b j If the spatial correlation of f(a*),f(a j )) will be small due to the Gaussian kernel in equation (4). From a modeling perspective, it is undesirable to train the Gaussian process model in equation (1) with samples that have small correlations, as this will degrade the model performance. Mathematically, this can be proven as follows.

[0070] For the query a* in Algorithm 2 and the calibrated Gaussian process model, the model prediction

number

number

[0071] Suppose,

number

number

number

number

number

number

number

number

number

number

number

[0072] In some embodiments, to ensure that both spatial and time related sample in D can contribute, a “space-time” JITL (ST-JITL) approach is used with the following space-time Raman model (e.g., as local model 132): b i =g(a i ,t i )+E i Formula (26) Where:

number

number

number

number

number

number

number

number

number

[0073] Unlike the variables a and b, the role of t in equations (30a) and (30b) is simply D T Note that the aim of this study is to improve the contribution of . Physically, given a, the variable t has no effect on b. So if

number

number

number

number

number

number

number

number

number

number

number

number

number

number

[0074] From equations (30a) and (30b), the covariance rθ K space and K time Given the covariance functions for the spatiotemporal Raman model in equations (30a) and (30b), it is easy to verify that the covariance functions include contributions from both the kernel parameters θ and the noise variance σ 2 can be estimated by maximizing

number

number

number

number

number

number

number

number

number

[0075] From the above, equations (38a) and (38b) still hold true for k space and k time It can be seen that the ST-JITL technique includes contributions from both. An example algorithm that formally outlines the ST-JITL technique is provided in Algorithm 3 below.

number

[0076] A-JITL and ST-JITL (in Algorithms 2 and 3, respectively) 1 Note that it may be the same when β 1 =0, K time For =0, r θ =k space =k θ (as can be seen from equations (28) and (29b)).

[0077] Turning now to FIG. 6, an exemplary data flow 350 is shown that may occur when analyzing a biopharmaceutical process using ST-JITL techniques as described herein. Data flow 350 may occur, for example, in system 100 of FIG. 1 or system 150 of FIG. 2. In data flow 350, spectral data 352 is provided by a spectrometer / probe. For example, spectral data 352 may include Raman scan vectors generated by Raman analyzer 106, or NIR scan vectors, etc. Query points 354 are generated (e.g., by query unit 140) based on the spectral data 352 and are used to query a global dataset 356 that may include, for example, all observational datasets in observation database 136. Global dataset 356 is logically separated into the last k entries 357A (e.g., all from the current experiment / process) and all entries 357B before the last k entries 357A (e.g., from previous and possibly from the current experiment / process). The value of k may be determined based on the number of samples at query point 354. Local dataset 358 may be generated from entries 357A and 357B according to Algorithm 3, for example.

[0078] The local data set 358 is then used as training data (e.g., by the local model generator 142) to calibrate a local model 360 (e.g., the local model 132). The local model 360 then calculates the media component concentrations, media conditions (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+ and other nutrients or metabolites, pH, pCO 2 , pO 2 , temperature, osmolality, etc.), viable cell density, titer, important quality attributes, cell condition, etc., and possibly also output confidence limits or another suitable confidence metric.

[0079] When an actual analytical measurement (e.g., a measurement made by an analytical instrument such as one of the analytical instruments 104) is available, a new entry 364 (including its sample number) is created and added to the global dataset 356. Such measurements may be available, for example, on a periodic sampling basis (e.g., once or twice a day) and / or may become available in response to a trigger having variable timing (e.g., if predictions for a certain number in a row have unacceptably wide confidence limits, etc.).

[0080] As described above, analytical measurements may be scheduled / triggered based on the current and / or recent performance of one or more local models (e.g., local models 132, 260, 310, or 360) to maintain or improve prediction accuracy while reducing resource usage (e.g., analytical equipment usage). This technique may be used, for example, in conjunction with A-JITL, ST-JITL, or straight JITL.

[0081] In one embodiment, the confidence interval is used to trigger model maintenance. In particular, if the width of the confidence interval (e.g., the distance between the confidence limits as calculated using equation (16) or equations (37a), (37b)) around a given model prediction (e.g., around the most recent prediction made by the local model 132, 260, 310, or 360) is greater than a predefined threshold, the database maintenance unit 146 may generate a request message and cause the computer 110 to send a message to the analytical instrument 104 to request a measurement. For example, in the example results of FIG. 3, the database maintenance unit 146 may trigger new analytical measurements near the end of 12 / 08 / 2017, 12 / 09 / 2017, and 12 / 14 / 2017, where the shaded portion 208 indicates the wide confidence interval (i.e., b U -b L (large values ​​of ).

[0082] In response to the request message, the analytical measurement 104 performs the measurement and provides the measurement to the computer 110. The database maintenance unit 146 may then transmit the measurement and the corresponding Raman scan vector received from the Raman analyzer 106 to the database server 112 for storage in the observation database 136. For example, the measurement and the scan vector may be added to library L (in the case of straight JITL) or library K (in the case of A-JITL or ST-JITL) discussed above.

[0083] Conversely, if the width of the confidence interval around a given model prediction is not greater than a predefined threshold, the database maintenance unit 146 may not request new analytical measurements, in which case the library in the observations database 136 remains unchanged. The analytical instrument 104 may also obtain information about media component concentrations, media conditions (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+ and other nutrients or metabolites, pH, pCO 2 , pO 2 In embodiments including multiple instruments measuring different characteristics such as cell density, temperature, osmolality, etc., viable cell density, titer, important quality attributes, cell condition, etc., and where separate local models are used to predict different characteristic values, the scheduling process may be performed separately for the analytical instruments measuring each predicted characteristic and possibly characteristics with different confidence interval width thresholds for each characteristic.

[0084] Mathematically, the database maintenance unit 146 may schedule / trigger a new analytical measurement at query point a* under the following conditions: b U -b L ≧THR Formula (39) Here, THR is a user-defined threshold. In some embodiments, THR may be adjusted by the user to suit a particular application or use case. For example, the user may set a relatively low THR value (used by the database maintenance unit 146) for applications where model reliability is important, thereby causing model / library maintenance operations to occur more frequently. Generally, THR may be set to different values based on process criticality, predicted parameters such as media component concentration, media state (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+ and other nutrients or metabolites, pH, pCO 2 , pO 2 , temperature, osmotic pressure, etc.), viable cell density, titer, important quality attributes, cell state, etc., and / or based on the current period (e.g., using a lower THR on later days of culture compared to the first day). The selection of THR represents a trade-off between model accuracy and resource (analytical instrument) usage, and lower thresholds tend to increase model accuracy at the expense of increased resource usage.

[0085] Modifications to this scheduling protocol are also possible. In one embodiment, for example, the database maintenance unit 146 may apply one or more model performance criteria not only to the current (latest) prediction but also to one or more other latest predictions (e.g., the latest N predictions where N > 1). As an example of such an embodiment, the database maintenance unit 146 may calculate the average width of the confidence intervals for the latest N predictions (N ≧ 1) and then compare that average width to the threshold THR. As another example, the database maintenance unit 146 may identify the X largest confidence intervals among the last Y predictions (X < Y) and schedule / trigger a new analytical measurement only if the width of each of those X intervals is greater than the threshold THR.

[0086] 7 is a flow diagram of an example method 400 for analyzing (e.g., for monitoring and / or controlling) a biopharmaceutical process. Method 400 may be implemented by computer 110 of FIG. 1 (e.g., by processing unit 120 executing instructions of JITL predictive application 130) or a computer of FIG. 2, and / or by a server, such as, for example, database server 112 of FIG. 1 or FIG. 2.

[0087] At block 402, query points associated with a scan of the biopharmaceutical process by a spectroscopy system (e.g., by the Raman analyzer 104 and Raman probe 106 of system 100 or system 150) are identified. The query points may be identified, for example, based at least in part on a spectral scan vector (e.g., Raman or NIR scan vector) generated by the spectroscopy system when scanning the biopharmaceutical process. Depending on the embodiment, the query points may be identified based on the raw spectral scan vector or after appropriate pre-processing filtering of the raw spectral scan vector. In some aspects, the query points are also identified based on a media profile associated with the biopharmaceutical process (e.g., fluid type, specific nutrients, pH levels, etc.) and / or other information, such as, for example, one or more operating conditions at which the biopharmaceutical process is analyzed (e.g., metabolite concentration set points, etc.).

[0088] At block 404, an observational database (e.g., observational database 136) is queried. The observational database may include observational data sets relating to past observations of multiple biopharmaceutical processes. Each of the observational data sets may include spectral data (e.g., Raman or NIR scan vectors) and a corresponding analytical measurement (or in some embodiments, more than one analytical measurement). The analytical measurements may include, for example, media component concentrations, media conditions (e.g., glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na+, K+ and other nutrients or metabolites, pH, pCO 2 , pO 2, temperature, osmolarity, etc.), viable cell density, titer, important quality attributes, and / or cell condition.

[0089] Block 404 may include selecting as training data those observation data sets that satisfy one or more relevance criteria with respect to the query point. If the query point includes a spectral scan vector, for example, block 404 may include comparing the spectral scan vector to the spectral scan vectors associated with each of the past observations represented in the observation database (e.g., by calculating a Euclidean or other distance between (1) the specific underlying spectral scan vector of the query point and (2) each of the spectral scan vectors associated with the past observations, and then selecting as training data any of the spectral scan vectors associated with the past observations identified as being within a threshold distance of the specific underlying spectral scan vector of the query point).

[0090] The selected training data is used to calibrate a local model specific to the biopharmaceutical process being monitored in block 406. The local model (e.g., local model 132) is trained in block 406 to predict analytical measurements based on the spectral data input (e.g., Raman or NIR spectral scan vectors). In some embodiments, the local model is a Gaussian process machine learning model.

[0091] In block 408, an analytical measurement of the biopharmaceutical process is predicted using the local model. Block 408 may include using the local model to analyze spectral data (e.g., Raman or NIR scan vectors) generated by the spectroscopy system in scanning the biopharmaceutical process. For example, block 408 may include using the local model to predict the analytical measurement by processing the same scan vectors on which the query points were based or other spectral data. Depending on the embodiment, the local model may be used to analyze the raw spectral data (e.g., raw Raman scan vectors) or to analyze the spectral data after appropriate pre-processing filtering of the raw spectral data. In some embodiments, block 408 also includes identifying a confidence metric (e.g., confidence limit, confidence score, etc.) associated with the predicted analytical measurement of the biopharmaceutical process. In some embodiments, the local model also predicts one or more additional analytical measurements in block 408.

[0092] In some embodiments, method 400 includes one or more additional blocks not shown in FIG. 5. For example, method 400 may include an additional block in which at least one parameter of a biopharmaceutical process is controlled based at least in part on the predicted analytical measurement in block 408. In some embodiments, the parameter may be of the same type as the predicted analytical measurement (e.g., controlling glucose concentration based on predicted glucose concentration) or may be of a different type. Model predictive control (MPC) techniques may be used, for example, to control the parameter (or multiple parameters).

[0093] As another example, method 400 may include a first additional block in which an actual analytical measurement of the biopharmaceutical process is obtained (e.g., by or from one of the analytical instruments 104 in response to identifying the predicted analytical measurement, and possibly one or more prior / recent measurements, as discussed above, as meeting / not meeting one or more model performance criteria), and a second additional block in which (1) the spectral data generated by the spectroscopy system when the actual analytical measurement is obtained, and (2) the actual analytical measurement of the biopharmaceutical process is added to the observation database (e.g., by sending the spectral data and analytical measurement to a database server, such as database server 112, or by directly adding the spectral data and analytical measurement to a local observation database, etc.). In embodiments in which multiple types of analytical measurements are predicted, multiple actual analytical measurements may be obtained and added to the observation database.

[0094] As yet another example, method 400 may include one or more sets of additional blocks, each similar to blocks 402-408. In each of these sets of additional blocks, the local models may be calibrated by querying the observational database (or other observational databases) and used to predict different types of analytical measurements.

[0095] Additional considerations pertaining to this disclosure are now addressed.

[0096] The terms "polypeptide" or "protein" are used interchangeably throughout and refer to molecules that contain two or more amino acid residues linked together by peptide bonds. Polypeptides and proteins also include polymers that have one or more deletions, insertions, and / or substitutions from the amino acid residues of the native sequence, i.e., polypeptides or proteins produced by naturally occurring non-recombinant cells, or molecules produced by genetically engineered or recombinant cells that have one or more deletions, insertions, and / or substitutions from the amino acid residues of the native protein's amino acid sequence. Polypeptides and proteins also include amino acid polymers in which one or more amino acids are chemical analogs of the corresponding naturally occurring amino acids and polymers. Polypeptides and proteins also include modifications, including, but not limited to, glycosylation, lipid conjugation, sulfation, gamma-carboxylation of glutamic acid residues, hydroxylation, and ADP-ribosylation.

[0097] Polypeptides and proteins may be of scientific or commercial interest, including protein-based therapeutics. Proteins include, inter alia, secreted proteins, non-secreted proteins, intracellular proteins, or membrane-bound proteins. Polypeptides and proteins may be produced by recombinant animal cell lines using cell culture techniques and may be referred to as "recombinant proteins." The expressed protein may be produced intracellularly or secreted into the culture medium from which it can be harvested and / or collected. Proteins include proteins that exert a therapeutic effect by binding to targets, particularly those listed below, including targets derived from, related to, and modifications thereof.

[0098] The protein is an "antigen binding protein." Antigen binding protein refers to a protein or polypeptide that contains an antigen binding region or portion that has a strong affinity for another molecule (antigen) to which it binds. Antigen binding proteins include antibodies, peptibodies, antibody fragments, antibody derivatives, antibody analogs, fusion proteins (including single chain variable fragments (scFv) and dual chain (bivalent) scFv), muteins, xMAbs, and chimeric antigen receptors (CARs).

[0099] An scFv is a single chain antibody fragment that has the variable regions of the heavy and light chains of an antibody linked together. See U.S. Patent Nos. 7,741,465 and 6,319,494 and Eshhar, et al., Cancer Immunol. Immunotherapy (1997) 45:131-136. An scFv retains the ability of the parent antibody to specifically interact with a target antigen.

[0100] The term "antibody" includes reference to both glycosylated and non-glycosylated immunoglobulins of any isotype or subclass, or antigen-binding regions thereof that compete with the intact antibody for specific binding. Unless otherwise specified, antibodies include human, humanized, chimeric, multispecific, monoclonal, polyclonal, hetero-IgG, XmAbs, bispecific, and oligomers or antigen-binding fragments thereof. Antibodies include lgG1, lgG2, lgG3, or lgG4 types. Also included are proteins having antigen-binding fragments or regions, such as Fab, Fab', F(ab')2, Fv, diabodies, Fd, dAb, maxibodies, single-chain antibody molecules, single domain VHH, complementarity determining region (CDR) fragments, scFv, diabodies, triabodies, tetrabodies, and polypeptides that contain at least a portion of an immunoglobulin sufficient to confer specific antigen binding to a target polypeptide.

[0101] Also included are human, humanized, and other antigen binding proteins, such as human and humanized antibodies, that do not produce a significant adverse immune response when administered to humans.

[0102] Also included are peptibodies, which are polypeptides comprising one or more biologically active peptides linked together, optionally via a linker, by an Fc domain, see U.S. Patent Nos. 6,660,843, 7,138,370, and 7,511,012.

[0103] Proteins also include engineered receptors such as chimeric antigen receptors (CARs or CAR-Ts) and T cell receptors (TCRs). CARs typically incorporate an antigen-binding domain (such as an scFv) in conjunction with one or more costimulatory ("signaling") domains and one or more activation domains.

[0104] Also included are bispecific T cell engager (BiTE®) antibody constructs, which are recombinant protein constructs made from two flexibly linked antibody-derived binding domains (see WO 99 / 54440 and WO 2005 / 040220). One binding domain of the construct is specific for a selected tumor-associated surface antigen on the target cell; the second binding domain is specific for CD3, a subunit of the T cell receptor complex on T cells. BiTE® constructs may also include the ability to bind to a context-independent epitope at the N-terminus of the CD3s chain (WO 2008 / 119567) to more specifically activate T cells. Half-life extended BiTE® constructs include fusion of small bispecific antibody constructs, preferably to larger proteins that do not interfere with the therapeutic effect of the BiTE® antibody construct. Examples of such further development of bispecific T cell engagers include the bispecific Fc molecules described, for example, in US2014 / 0302037, US2014 / 0308285, WO2014 / 151910, and WO2015 / 048272. Another strategy is the use of human serum albumin (HAS) fused to the bispecific molecule, or simply the fusion of a human albumin binding peptide (see, for example, WO2013 / 128027, WO2014 / 140358). Another HLE BiTE® strategy involves fusing a first domain that binds to a target cell surface antigen, a second domain that binds to an extracellular epitope of the human and / or rhesus CD3e chain, and a third domain that is a specific Fc modality (WO 2017 / 134140).

[0105] Also included are modified proteins, such as proteins that are chemically modified by non-covalent, covalent, or both covalent and non-covalent bonds, and also proteins that further include one or more post-translational modifications, which may be made by cell engineering systems, or modifications introduced ex vivo by enzymatic and / or chemical methods, or otherwise introduced.

[0106] Proteins may also include recombinant fusion proteins that contain multimerization domains such as, for example, leucine zippers, coiled coils, Fc portions of immunoglobulins, etc. Also included are proteins that contain all or part of the amino acid sequence of a differentiation antigen (referred to as CD proteins) or their ligands, or proteins generally similar to any of these.

[0107] In some embodiments, the protein may comprise a colony stimulating factor, such as granulocyte colony stimulating factor (G-CSF). Such G-CSF agents include, but are not limited to, Neupogen® (filgrastim) and Neulasta® (pegfilgrastim). Also included are Epogen® (epoetin alfa), Aranesp® (darbepoetin alfa), Dynepo® (epoetin delta), Mircera® (methoxypolyethylene glycol-epoetin beta), Hematide®, MRK-2578, INS-22, Retacrit® (epoetin zeta), Neorecormon® (epoetin beta), Silapo® (epoetin zeta), Binocrit® (epoetin alfa), epoetin alfa H Also included are, but are not limited to, exal, Abseamed® (epoetin alfa), Ratioepo® (epoetin theta), Eporatio® (epoetin theta), Biopoin® (epoetin theta), epoetin alfa, epoetin beta, epoetin zeta, epoetin theta, and epoetin delta, epoetin omega, epoetin iota, tissue plasminogen activator, GLP-1 receptor agonists, and molecules or variants or analogs thereof and biosimilars of any of the foregoing.

[0108] In some embodiments, the proteins may include proteins that specifically bind to one or more CD proteins, HER receptor family proteins, cell adhesion molecules, growth factors, nerve growth factors, fibroblast growth factors, transforming growth factors (TGFs), insulin-like growth factors, osteoinductive factors, insulin and insulin-related proteins, coagulation and coagulation-related proteins, colony stimulating factors (CSFs), other blood and serum proteins, blood group antigens; receptors, receptor-related proteins, growth hormones, growth hormone receptors, T cell receptors; neurotrophic factors, neurotrophins, relaxins, interferons, interleukins, viral antigens, lipoproteins, integrins, rheumatoid factors, immunotoxins, surface membrane proteins, transport proteins, homing receptors, addressins, regulatory proteins, and immunoadhesins.

[0109] In some embodiments, the protein may include proteins that bind to one or more of the following, alone or in any combination: CD proteins, including but not limited to CD3, CD4, CD5, CD7, CD8, CD19, CD20, CD22, CD25, CD30, CD33, CD34, CD38, CD40, CD70, CD123, CD133, CD138, CD171, CD174, HER receptor family proteins, including HER2, HER3, HER4, EGF receptor EGFRvIII, cell adhesion molecules, e.g. growth factors, including, but not limited to, vascular endothelial growth factor ("VEGF"), VEGFR2, growth hormone, thyroid stimulating hormone, follicle stimulating hormone, luteinizing hormone, growth hormone releasing factor, parathyroid hormone, Mullerian inhibitory substance, human macrophage inflammatory protein (MIP-1-α), erythropoietin (EPO), nerve growth factors such as NGF-β, platelet derived growth factors (PDGF), including, for example, aFGF and bFGF fibroblast growth factor, epidermal growth factor (EGF), crypto, transforming growth factors (TGFs) including TGF-α and TGF-β, including TGF-β1, TGF-β2, TGF-β3, TGF-β4, or TGF-β5, insulin-like growth factors-I and -II (IGF-I and IGF-II), des(1-3)-IGF-I (brain IGF-I), and bone morphogenetic factors, insulin and insulin-related proteins including, but not limited to, insulin, insulin A chain, insulin B chain, proinsulin, and insulin-like growth factor binding protein. Other blood and serum proteins, including, but not limited to, proteins, coagulation and coagulation-related proteins such as factor VIII, tissue factor, von Willebrand factor, among others, plasminogen activators such as protein C, alpha-1-antitrypsin, urokinase, and tissue plasminogen activator ("t-PA"), bombadin, thrombin, thrombopoietin, thrombopoietin receptors, colony stimulating factors (CSFs), including M-CSF, GM-CSF, and G-CSF, among others, albumin, IgE, and blood group antigens, e.g.,(x) receptors and receptor-associated proteins, including flk2 / flt3 receptor, obesity (OB) receptor, growth hormone receptor, and T cell receptor; (x) neurotrophic factors, including, but not limited to, bone-derived neurotrophic factor (BDNF) and neurotrophin-3, -4, -5, or -6 (NT-3, NT-4, NT-5, or NT-6); (xi) relaxin A chain, relaxin B chain, and prorelaxin, interferons, including, for example, interferon-α, -β, and -γ, interleukins (IL), for example, IL-1 to IL-10, IL-12, IL-15, IL- 17, IL-23, IL-12 / IL-23, IL-2Ra, IL1-R1, IL-6 receptor, IL-4 receptor and / or IL-13RA2 which is a receptor for IL-13, or IL-1RAP which is an IL-17 receptor; (xiv) viral antigens including, but not limited to, AIDS envelope virus antigen, lipoproteins, calcitonin, glucagon, atrial natriuretic factor, pulmonary surfactant, tumor necrosis factor-α and -β, enkephalinase, BCMA, IgKappa, ROR-1, ERBB2, mesothelin, RANTES (Regulated Antigens of Escherichia coli), and the like. on activation normal T-cell expressed and secreted), mouse gonadotropin-related peptide, Dnase, FR-α, inhibin, and activin, integrin, protein A or D, rheumatoid factor, immunotoxin, bone morphogenetic protein (BMP), superoxide dismutase, surface membrane protein, decay accelerating factor (DAF), AIDS envelope, transport protein, homing receptor, MIC (MIC-a, MIC-B), ULBP 1-6, EPCAM, addressin, regulatory protein, immunoadhesin, antigen-binding protein, somatropin, CTGF, CTLA4, eotaxin-1, MUC1, CEA, c-MET, claudin-18, GPC-3, EPHA2, FPA, LMP1, MG7, NY-ESO-1, PSCA, ganglioside GD2, ganglioside GM2, BAFF, OPGL (RANKL), myostatin, Dickkopf-1 (DKK-1), Ang2, NGF, IGF-1 receptor, hepatocyte growth factor (HGF), TRAIL-R2, c-Kit, B7RP-1, PSMA, NKG2D-1,Programmed cell death protein 1 and ligand, PD1 and PDL1, mannose receptor / hCGβ, hepatitis C virus, mesothelin dsFv [PE38 conjugate, Legionella pneumophila (lly), IFN gamma, interferon gamma-inducible protein 10 (IP10), IFNAR, TALL-1, thymic stromal lymphopoietin (TSLP), stem cell factor, Flt-3, calcitonin gene-related peptide (CGRP), OX40L, α4β7, platelet specific (platelet glycoprotein Iib / IIIb (PAC-1), transforming growth factor β (TFGβ), zona pellucida sperm-binding protein 3 (ZP-3), TWEAK, platelet-derived growth factor receptor α (PDGFRα), sclerostin, and biologically active fragments or variants of any of the foregoing.

[0110] In another embodiment, the protein is selected from the group consisting of abciximab, adalimumab, adecatumumab, aflibercept, alemtuzumab, alirocumab, anakinra, atacicept, basiliximab, belimumab, bevacizumab, biosuzumab, blinatumomab, brentuximab vedotin, brodalumab, cantuzumab mertansine, canakinumab, cetuximab, certolizumab pegol, conatumumab, daclizumab, denosumab, eculizumab, edrecolomab, efalizumab, epratuzumab, etanercept, evolocumab, galiximab, ganitumab, gemtuzumab, golimumab, ibritumomab tiuxetan, infliximab, ipilimumab , lerdelimumab, rumiliximab, ixekizumab (lxdkizumab), mapatuzumab, motesanib diphosphate, muromonab-CD3, natalizumab, nesiritide, nimotuzumab, nivolumab, ocrelizumab, ofatumumab, omalizumab, oprelvekin, palivizumab, panitumumab, pembrolizumab, pembrolizumab, Includes nitumumab, pexelizumab, ranibizumab, rilotumumab, rituximab, romiplostim, romosozumab, sargramostim, tocilizumab, tositumomab, trastuzumab, ustekinumab, vedolizumab, visilizumab, volociximab, zanolimumab, zalutumumab, and biosimilars of any of the above.

[0111] Proteins encompass all of the above, and further include antibodies comprising one, two, three, four, five, or six of the complementarity determining regions (CDRs) of any of the above antibodies. Also included are variants that comprise regions that are 70% or more, particularly 80% or more, more particularly 90% or more, even more particularly 95% or more, particularly 97% or more, more particularly 98% or more, and even more particularly 99% or more amino acid sequence identical to the reference amino acid sequence of the subject protein. Identity in this regard can be determined using a variety of well-known and readily available amino acid sequence analysis software. Preferred software includes those that implement the Smith-Waterman algorithm, which is considered a satisfactory solution to the problem of searching and aligning sequences. Other algorithms may also be employed, especially where speed is an important consideration. Commonly used programs for alignment and homology matching of DNA, RNA, and polypeptides that can be used in this regard include FASTA, TFASTA, BLASTN, BLASTP, BLASTX, TBLASTN, PROSRCH, BLAZE, and MPSRCH, the latter of which is an implementation of the Smith-Waterman algorithm for running on massively parallel processors produced by MasPar.

[0112] Some of the figures described herein show example block diagrams having one or more functional components. It should be understood that such block diagrams are for illustrative purposes, and that the devices described and shown may have more, fewer, or alternative components than those shown. Additionally, in various embodiments, the components (as well as the functionality provided by individual components) may be associated with or otherwise integrated as part of any suitable component.

[0113] Embodiments of the present disclosure relate to non-transitory computer-readable storage media having computer code for performing various computer-implemented operations. The term "computer-readable storage medium" is used herein to include any medium capable of storing or encoding sequences of instructions or computer code for performing the operations, methods, and techniques described herein. The media and computer code may be those specially designed and constructed for the purposes of the embodiments of the present disclosure, or they may be of the type well known and available to those skilled in the art of computer software technology. Examples of computer-readable storage media include, but are not limited to, magnetic media such as hard disks, floppy disks, and magnetic tapes, optical media such as CD-ROMs and holographic devices, magneto-optical media such as optical disks, and hardware devices specially configured to store and execute program code, such as ASICs, programmable logic devices ("PLDs"), and ROM and RAM devices.

[0114] Examples of computer code include machine code, such as that produced by a compiler, and files containing higher level code executed by a computer using an interpreter or compiler. For example, an embodiment of the present disclosure may be implemented using Java, C++, or other object-oriented programming language and development tools. Further examples of computer code include encrypted and compressed code. Furthermore, an embodiment of the present disclosure may be downloaded as a computer program product, which may be transferred from a remote computer (e.g., a server computer) to a requesting computer (e.g., a client computer or a different server computer) over a transmission channel. Other embodiments of the present disclosure may be implemented as hardwired circuitry in place of, or in combination with, machine-executable software instructions.

[0115] As used herein, the singular terms "a," "an," and "the" may include plural references unless the context clearly dictates otherwise.

[0116] As used herein, the terms "connect," "connected," and "connection" mean an operative coupling or linking. Connected components may be directly or indirectly coupled to each other, for example through a set of separate components.

[0117] As used herein, the terms "approximately," "substantially," "substantial," and "about" are used to describe and explain small variations. When used in the context of an event or circumstance, these terms can refer to instances where the event or circumstance occurs exactly, as well as instances where the event or circumstance occurs to a close degree. For example, when used in the context of a numerical value, these terms can refer to a range of variation of ±10% or less of the numerical value, such as ±5% or less, ±4% or less, ±3% or less, ±2% or less, ±1% or less, ±0.5% or less, ±0.1% or less, or ±0.05% or less. For example, two numerical values ​​may be determined to be "substantially" identical if the difference between the values ​​is within ±10%, such as within ±5%, ±4%, ±3%, ±2%, ±1%, ±0.5%, ±0.1%, or ±0.05% of the mean of the values.

[0118] In addition, quantities, ratios, and other numerical values ​​are sometimes presented herein in a range format. This range format is used for convenience and brevity, and while it includes numerical values ​​explicitly stated as range limits, it is understood that these should be flexibly read as including all individual numerical values ​​or subranges subsumed within that range, as if each numerical value and subrange was expressly stated.

[0119] Although the present disclosure has been described and illustrated with reference to specific embodiments thereof, these descriptions and illustrations are not intended to limit the present disclosure. It should be understood by those skilled in the art that various modifications may be made and equivalents may be substituted without departing from the true spirit and scope of the present disclosure as defined by the appended claims. The illustrations may not necessarily be to scale. Due to manufacturing processes, tolerances, and / or other reasons, there may be differences between the technical expressions in the present disclosure and the actual devices. There may be other embodiments of the present disclosure that are not specifically shown. The present specification and drawings (other than those claimed) should be regarded as illustrative and not limiting. Modifications may be made to adapt a particular situation, material, composition of matter, technique, or process to the objective, spirit, and scope of the present disclosure. All such modifications are intended to be within the scope of the claims appended hereto. Although the techniques disclosed herein are described with reference to particular operations being performed in a particular order, it should be understood that these operations may be combined, sub-divided, or re-ordered to form equivalent techniques without departing from the teachings of this disclosure. Thus, unless specifically indicated herein, the order and grouping of operations is not intended to be a limitation of this disclosure.

Claims

1. 1. A computer-implemented method for monitoring and / or controlling a biopharmaceutical process, comprising: identifying, by one or more processors, query points associated with a scan of the biopharmaceutical process by a spectroscopy system; querying, by the one or more processors, an observational database including a plurality of observational data sets related to past observations of a biopharmaceutical process, each of the observational data sets including spectral data and corresponding actual analytical measurements, wherein querying the observational database includes selecting, from the plurality of observational data sets, observational data sets that satisfy one or more relevance criteria with respect to the query points as training data; calibrating, by the one or more processors and with the selected training data, a local model specific to the biopharmaceutical process, wherein the local model is trained to predict analytical measurements based on spectral data input; predicting, by the one or more processors, an analytical measurement of the biopharmaceutical process, where predicting the analytical measurement of the biopharmaceutical process includes analyzing spectral data generated by the spectroscopy system while scanning the biopharmaceutical process with the local model. method.

2. a) the spectroscopy system is a Raman spectroscopy system; and b) identifying query points includes identifying the query points based at least in part on a spectral scan vector, the spectral scan vector being generated by the spectroscopy system when scanning the biopharmaceutical process; selecting as training data the observational data sets that satisfy one or more relevance criteria with respect to the query points comprises comparing the spectral scan vectors underlying the identification of the query points with spectral scan vectors associated with the past observations of the biopharmaceutical process; The computer-implemented method of claim 1 , comprising at least one of:

3. Identifying a query point further includes identifying the query point based on a sample number associated with the spectral scan vector; Selecting as training data the observational data sets that satisfy one or more relevance criteria with respect to the query point includes: (i) comparing the spectral scan vector on which the query point was identified with spectral scan vectors associated with the past observations of the biopharmaceutical process; and (ii) comparing the number of samples associated with the query point with the number of samples associated with the past observations of the biopharmaceutical process.

3. The computer-implemented method of claim 2.

4. Selecting as training data the observation data set that satisfies one or more relevance criteria with respect to the query points comprises: selecting a most recent k observation data set for inclusion in the training data; 4. The computer-implemented method of claim 3.

5. a) predicting the analytical measurement of the biopharmaceutical process includes analyzing the spectral scan vectors underlying the identification of the query points using the local model; and b) selecting as training data the observation data sets that satisfy one or more relevance criteria with respect to the query points, (i) calculating a distance between the spectral scan vector underlying the identification of the query point and (ii) the spectral scan vector associated with the past observations of the biopharmaceutical process; selecting as the training data any of the spectral scan vectors associated with the past observations that are within a threshold distance of the spectral scan vector on which the query point was identified; The computer-implemented method of any one of claims 2 to 4, further comprising at least one of:

6. a) identifying the query points includes identifying the query points based at least in part on one or both of: (i) a media profile associated with the biopharmaceutical process; and (ii) one or more operating conditions under which the biopharmaceutical process is analyzed; b) calibrating the local model specific to the biopharmaceutical process comprises calibrating a Gaussian process machine learning model specific to the biopharmaceutical process; c) calibrating the local model specific to the biopharmaceutical process includes calibrating a model that is a function of both the spectral data and the number of samples in a given observational data set; d) predicting an analytical measurement of the biopharmaceutical process includes identifying a confidence index associated with the predicted analytical measurement of the biopharmaceutical process using the local model; e) the method further comprising controlling, by the one or more processors and based at least in part on the predicted analytical measurement of the biopharmaceutical process, at least one parameter of the biopharmaceutical process; f) the predicted analytical measurement of the biopharmaceutical process is a media component concentration, a media condition, a viable cell density, a titer, a critical quality attribute, or a cell condition; g) the predicted analytical measurements of the biopharmaceutical process are glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na ＋ , or K ＋ The concentration is h) The predicted analytical measurements of the biopharmaceutical process include pH, pCO ２ , pO ２ , temperature, or osmolarity; i) the method further includes acquiring, by an analytical instrument, an actual analytical measurement of the biopharmaceutical process, and causing the one or more processors to add to the observations database: (i) spectral data generated by a spectroscopy system when the actual analytical measurement is acquired, and (ii) the actual analytical measurement of the biopharmaceutical process; and j) said biopharmaceutical process is a cell culture process; The computer-implemented method of any one of claims 1 to 5, comprising at least one of:

7. identifying, by the one or more processors, at least the predicted analytical measurements as not meeting one or more model performance criteria; obtaining the actual analytical measurement is performed in response to identifying at least the predicted analytical measurement as not meeting the one or more model performance criteria.

7. The computer-implemented method of claim 6.

8. Identifying at least the predicted analytical measurement as not meeting the one or more model performance criteria includes: generating a confidence interval associated with the predicted analytical measurement; comparing the confidence interval to a predefined threshold.

8. The computer-implemented method of claim 7.

9. 1. A spectroscopic system for monitoring and / or controlling a biopharmaceutical process, comprising: one or more spectroscopic probes collectively configured to: (i) deliver a source electromagnetic radiation to the biopharmaceutical process; and (ii) collect electromagnetic radiation while the source electromagnetic radiation is delivered to the biopharmaceutical process; one or more memories that collectively store an observational database including a plurality of observational data sets relating to past observations of a biopharmaceutical process, each of the observational data sets including spectral data and corresponding actual analytical measurements; One or more processors, identifying query points associated with scanning the biopharmaceutical process with the spectroscopic system; Querying the observational database by at least selecting as training data from among the plurality of observational data sets those observational data sets that satisfy one or more relevance criteria with respect to the query points; calibrating a local model specific to the biopharmaceutical process using the selected training data, the local model being trained to predict analytical measurements based on spectral data input; and one or more processors configured to predict, using at least the local model, an analytical measurement of the biopharmaceutical process by analyzing spectral data generated by the spectroscopy system as the one or more spectroscopic probes scan the biopharmaceutical process. Spectroscopic system.

10. a) the spectroscopy system is a Raman spectroscopy system; b) the one or more processors; identifying the query points based at least in part on a spectral scan vector, the spectral scan vector being generated by the spectroscopy system when scanning the biopharmaceutical process; selecting said training data by at least comparing said spectral scan vectors on which said query points were identified with spectral scan vectors associated with said past observations of said biopharmaceutical process; c) the local model is a Gaussian process machine learning model; d) the local model is a function of both the spectral data and the number of samples in a given observational data set; e) the one or more processors are further configured to use the local model to determine a confidence index associated with the predicted analytical measurement of the biopharmaceutical process; f) the one or more processors are further configured to control at least one parameter of the biopharmaceutical process based at least in part on the predicted analytical measurement of the biopharmaceutical process; g) the predicted analytical measurement of the biopharmaceutical process is a media component concentration, a media condition, a viable cell density, a titer, a critical quality attribute, or a cell condition; h) the predicted analytical measurements of the biopharmaceutical process are glucose, lactate, glutamate, glutamine, ammonia, amino acids, Na ＋ , or K ＋ The concentration is i) the predicted analytical measurements of the biopharmaceutical process include pH, pCO ２ , pO ２ , temperature, or osmolarity; j) an analytical instrument configured to obtain actual analytical measurements of the biopharmaceutical process; the one or more processors are further configured to: (i) cause spectral data generated by a spectroscopy system when the actual analytical measurements are obtained, and (ii) cause the actual analytical measurements of the biopharmaceutical process to be added to the observations database; k) said one or more processors; Identifying at least the predicted analytical measurements as not meeting one or more model performance criteria; and further configured, in response to identifying at least the predicted analytical measurement as not meeting the one or more model performance criteria, to obtain the actual analytical measurement from the analytical instrument. l) said one or more processors comprising at least generating a confidence interval associated with the predicted analytical measurement; and comparing said confidence interval with a predefined threshold. configured to identify at least the predicted analytical measurements as not meeting the one or more model performance criteria; and m) said biopharmaceutical process is a cell culture process; The spectroscopy system of claim 9 , comprising at least one of:

11. The one or more processors: identifying the query points based in part on a sample number associated with the spectral scan vector; and (ii) selecting as training data those observational data sets that satisfy one or more relevance criteria with respect to the query points, in part, by comparing the spectral scan vectors on which the identification of the query points was based with spectral scan vectors associated with the past observations of the biopharmaceutical process; and (iii) comparing the sample counts associated with the query points with sample counts associated with the past observations of the biopharmaceutical process. The spectroscopy system of claim 10.

12. The one or more processors: by selecting the most recent k observation data sets for inclusion in the training data; The spectroscopic system of claim 11 , configured to select as training data the observational data sets that partially satisfy one or more relevance criteria with respect to the query points.

13. A non-transitory computer readable medium storing instructions for monitoring and / or controlling a biopharmaceutical process, the instructions, when executed by one or more processors, causing the one or more processors to perform the method of any one of claims 1 to 8.

14. 1. A bioreactor system comprising: a bioreactor chamber configured to contain a biopharmaceutical process; The spectroscopic system according to any one of claims 9 to 12, Bioreactor system.

15. 15. The bioreactor system of claim 14 configured to produce a recombinant protein in a cell culture process.

Images