Predict bioreactor product yields based on independent or multivariate analysis of multiple physical attributes
By using multivariate prediction models in bioreactors, predicting cell biomass and product density based on multiple sensor data, the problem of low sample contamination and information update frequency in the prior art is solved, and sample-free manufacturing and dynamic process control are achieved , improves the productivity and safety of cell therapy.
Patent Information
- Application Number
- CN201980029397.8
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Priority Date
- 2018-05-01
- Filing Date
- 2019-04-30
- Publication Date
- 2024-12-03
- Estimated Expiration
- 2039-04-30
AI Technical Summary
Existing bioreactor systems require physical invasion when determining cell biomass yield, resulting in risk of sample contamination, high cost, and low frequency of information updates, limiting the effective manufacturing of therapeutic cell therapies and patient safety.
By measuring multiple physical properties using multiple sensors in a bioreactor, a multivariate prediction model is established, the density or quantity of surviving cells and cell products is predicted, and process parameters are adjusted to achieve sample-free manufacturing and dynamic process control.
Reduces the risk of contamination, improves the reliability and productivity of cell products, reduces costs, achieves more frequent process decisions and higher availability of therapeutic doses, and improves patient safety.
Smart Images

Figure CN112041424B_ABST
Abstract
Description
[0001] Cross-Reference to Prior Applications
[0002] This application claims priority to U.S. Provisional Patent Application No. 62 / 665,155, filed on May 1, 2018, and entitled “PREDICTING BIOREACTOR PRODUCT PRODUCTION BASED ON INDEPENDENT OR MULTIVARIATE ANALYSIS OF MULTIPLE PHYSICALATTRIBUTES,” the contents of which are incorporated herein by reference in their entirety.
[0003] Field of this specification
[0004] The present invention relates to manufacturing processes for therapeutic cell technology in bioreactors. In particular, the present invention relates to multivariate predictive models for viable cell biomass.
[0005] Background of this manual
[0006] Bioreactors have a long history in the culture of cells and cell products. In its simplest form, a bioreactor is a container containing living cells and a matrix that is converted to a product over time. The product can be intracellular or extracellular, or in the case of therapeutic cell technology, the cells themselves. Bioreactors can include sensor technology that is designed to measure a single condition or multiple conditions within the culture medium matrix. Traditionally, these measured conditions can include (but are not limited to): temperature, pH, dissolved oxygen (DO), reduction-oxidation (redox) potential, agitation, airflow (in and out), pressure, and vessel weight. Advances in sensor technology have expanded the range of analyses that may include (but are not limited to): NIR, FTIR, Raman spectroscopy, optical, immobilized enzyme-linked metabolite measurements, radar, conductivity, refractometry, etc.
[0007] Cultivation of therapeutic cell technologies in fully closed, process-controlled bioreactors is a new manufacturing endeavor, with only two autologous products on the market as of 2017: Novartis’ Kymriah and Gilead’s Yescarta. However, there are many candidates in development. The selection and use of various sensor technologies in manufacturing bioreactor platforms will vary based on cost, reproducibility, reliability, ease of use, and safety.
[0008] In the case of existing bioreactor systems, a physically invasive procedure is employed to determine whether the production of cell biomass is increasing or otherwise as expected: authorized laboratory personnel are required to wear appropriate protective equipment (e.g., which may include gloves and gowns or fully encapsulated body suits), extract one or more aliquots of the culture medium matrix from the bioreactor, and then analyze the matrix contents to determine cell counts and other variables. This method of determining cell biomass production has certain disadvantages, including: 1) Removal of aliquots of the culture medium matrix is equivalent to removing cells that might otherwise have been used for therapeutic applications; 2) Physical removal of samples introduces the risk of contamination (which can result in the destruction of batches, losses of hundreds of thousands or even millions of dollars, and loss of life); 3) The requirements for personnel and equipment to physically obtain samples have associated time and monetary costs; and 4) Due to the disadvantages mentioned above, the number of sampling events is limited, typically to once or twice per day, providing limited information about the status of the cell culture and making it impossible to take effective corrective measures. For example, autologous cell therapies require patient-specific starting materials and are created to be administered only to the donor. Since the starting material and product output are very limited and single supply, the reduction of the number of viable therapeutic cells caused by contamination events and / or sample extraction will put patients at risk. Allogeneic cell therapy is not patient-specific and therefore represents an opportunity to expand the manufacture of multiple doses in a single manufacturing run. Larger volumes and related material costs represent a huge risk of loss for manufacturers experiencing contamination events based on sample extraction. From the perspective of (10 to 1000) patients, the risk range of contaminated production runs will range from delayed receipt of curative cell therapy to catastrophic loss of life caused by administration of damaged cell material.
[0009] It would be desirable to obviate or mitigate one or more of the above disadvantages.
[0010] Overview of this manual
[0011] Embodiments described herein include a combination of sensor-based data from a bioreactor to control various aspects of a manufacturing process, and to predict the density of viable cells and cell products. The predicted density or amount of viable cells and cell products is generated by utilizing a combination of relationships between multiple continuous process variables including physical attributes. These growth predictions based on measurements of multiple physical attributes are then used to inform process decisions (e.g., increasing feed, gas, or agitation rates).
[0012] Embodiments described herein contemplate providing more robust predictions by using multiple physical property sensor readings, allowing for redundant coverage in the event of sensor failure or anomalies and / or improved predictions based on multiple sensor inputs.
[0013] The embodiments described herein tend to allow sample-free manufacturing of cells and cell products, thereby reducing contamination risk and product loss from sample extraction.
[0014] In a first aspect, a computer-implemented method for approximating a product yield in a bioreactor is provided. The method includes: providing a bioreactor containing living cells in a matrix, the bioreactor being used to culture a product derived from the cells or a product of the cells during a manufacturing process for a period of time; providing two or more sensors for measuring corresponding two or more different physical properties of the matrix during the period of time; receiving sensor data from the two or more sensors at a processor; and determining an approximate amount of the product via the processor based on a predetermined or recursive algorithm that relates the sensor data to the product yield by considering the sensor data of the two or more different physical properties independently or by considering the sensor data of the two or more different physical properties by multivariate. In another aspect, a method for determining a supply rate of a process input to a bioreactor is provided. The method includes: providing a target product density; determining an approximate amount of the product via the above-mentioned computer-implemented method for approximating a product yield in a bioreactor; comparing the approximate amount of the product with the target product density via the processor; based on the comparison, determining via the processor a supply rate of the process input to the bioreactor required to achieve the target product density within the period of time; and adjusting the supply rate of the process input to the bioreactor to the determined supply rate of the process input.
[0015] In another aspect, a system for approximating the yield of a product in a bioreactor is provided. The system includes: a bioreactor for culturing a product derived from living cells in a matrix or a product of living cells in a matrix during a manufacturing process for a period of time; two or more sensors for measuring corresponding two or more different physical properties of the matrix during the period of time; a computing device, which is communicatively connected to the two or more sensors. The computing device includes: a processor; a memory; and a communication module. The processor is communicatively connected to the memory and the communication module and is configured to: receive sensor data from two or more sensors; and determine an approximate amount of the product by independently considering the sensor data of the two or more different physical properties or by multivariately considering the sensor data of the two or more different physical properties based on a predetermined or recursive algorithm stored in the memory that associates the sensor data with the product yield. In another aspect, a system for determining a supply rate of a process input to a bioreactor is provided. The system includes the system for determining an approximate amount of a product as described above, and a feedback controller communicatively coupled to a computing device and a bioreactor, wherein the processor is further configured to: compare the approximate amount of the product with a target product density; based on the comparison, determine a supply rate of a process input to the bioreactor required to achieve the target product density within the time period; and transmit the adjusted supply rate of the process input to the bioreactor to the feedback controller via the communication module. Upon receiving the adjusted supply rate of the process input from the communication module of the computing device, the feedback controller adjusts the supply rate of the process input to the bioreactor to the supply rate of the process input determined by the processor.
[0016] In yet another aspect, a non-transitory computer-readable medium is provided, comprising computer-executable instructions for approximating a product yield in a bioreactor for culturing a product derived from living cells in a matrix or a product of living cells in a matrix during a manufacturing process for a period of time, the bioreactor comprising two or more sensors for measuring corresponding two or more different physical properties of the matrix during the period of time. The computer-executable instructions, when executed, cause a processor to: determine an approximate amount of the product by considering the sensor data of the two or more different physical properties independently or by multivariately considering the sensor data of the two or more different physical properties based on a predetermined or recursive algorithm that correlates sensor data received from the two or more sensors with the product yield. BRIEF DESCRIPTION OF THE DRAWINGS
[0018] Exemplary embodiments will now be described with reference to the accompanying drawings, in which:
[0019] Figure 1Depicted is an example of a predictive model for pH versus time in a 10-L pluripotent stem cell bioreactor expansion (right) and process data with model fit (left).
[0020] Figure 2 Depicted are examples of a predictive model for cell count versus time in a 10 L pluripotent stem cell bioreactor expansion (right) and sample-based data with model fit (left);
[0021] Figure 3 An exemplary prediction model (right) and time-matched model fit (left) are depicted;
[0022] Figure 4 Describes the Figure 3 The equations shown in are applied to examples of continuous process pH data, sample-based cell counting, and cell mass prediction;
[0023] Figure 5 Depicted are exemplary continuous dissolved oxygen process data from process values (PV, top) and control values (CV, bottom);
[0024] Figure 6 depicts exemplary data for feed control achieved through biomass modeling;
[0025] Figure 7 Depicted are exemplary model fits for continuous DO data (left), continuous pH data (center), and sample-based cell counts (right);
[0026] Figure 8 Depicted is an example graph showing the percentage of difference between multivariate cell quality predictions and actual measurements, with the open series (combined multivariate and 5 prediction means) shown on the secondary axis;
[0027] Fig. 9 depicts an exemplary interface according to an exemplary embodiment;
[0028] Fig.10 A flowchart depicting an exemplary embodiment of the computer executable instructions described herein;
[0029] Fig.11 A flowchart depicting another exemplary embodiment of the computer executable instructions described herein;
[0030] Fig.12 A flowchart depicting another exemplary embodiment of the computer executable instructions described herein;
[0031] Fig.13 a flowchart depicting yet another exemplary embodiment of the computer-executable instructions described herein; and
[0032] Fig.14 A flow diagram depicts an exemplary embodiment of the system described herein.
[0033] Detailed Description of Non-Limiting Exemplary Embodiments
[0034] Unless defined otherwise, all technical and scientific terms used herein generally have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
[0035] Embodiments described herein are expected to provide dynamic process control and product harvest timing in a sample-free manner. The expected advantages of sample-free manufacturing of therapeutic cell technology can be achieved, including reduced contamination risk and the increase in the number of cells that can be used for subsequent treatment doses, which is beneficial to manufacturers and final patients. Embodiments described herein are expected to use online sensor technology to establish multivariate models to predict viable cell concentrations and allow dynamic process control and product harvest timing. In this way, embodiments described herein are expected to overcome the limitation of a single input single output control loop by combining data streams from multiple sources to predict a single output. By using multiple inputs, it is expected that embodiments described herein produce a more robust predictive algorithm that can withstand disturbances to the system and contribution data streams while continuously predicting the precise or closely approximate concentration of cells or cell products or related measurements.
[0036] Allogeneic Example – Pluripotent Stem Cell Manufacturing
[0037] In the manufacture of allogeneic pluripotent stem cells (PSCs), bioreactors can be used for scaled production to make large batches with multiple doses. In order to achieve clinically relevant cell yields, a seed train of process-controlled stirred tank bioreactors can be used. In this example, the correlation between pH and DO was demonstrated to predict viable cell density and provide the information needed to adaptively adjust the medium feed rate and predict the harvest time required to pass to the next container in the seed train.
[0038] As expansion of the PSC continues, waste products such as lactate accumulate, and nutrients / heat-labile components such as growth factors decrease. Culture medium feed solutions are typically applied to counteract these phenomena, and the rate is typically increased in response to cell growth and / or metabolite accumulation / consumption. Regardless of the variables driving the decision, feed rates are traditionally adjusted in response to some sample-based data. The inventors have discovered that previous process development runs with sampling have established an optimal cell density at which feed rates were increased, and have developed a correlation between pH and DO and viable cell density that allows feed rates to be adjusted without drawing samples.
[0039] Process correlations are developed by fitting the continuous process data with an appropriate function. In one example, the data fit well (r2 = 0.976) to a logistic 4-parameter Hill function, but those skilled in the art will appreciate that there are many other models that can be utilized. The equations and fits to the raw data are given in Figure 1 It is shown in Figure 1 Depicted are examples of a predictive model for pH versus time in a 10-L pluripotent stem cell bioreactor expansion (right) and process data with model fit (left).
[0040] In another manufacturing run, samples were drawn and cells were counted. This allowed a fitted growth curve to be generated. In this case, the Gompertz 3 parameter equation fit well (r2 = 0.991), but those skilled in the art will appreciate that there are many other models that can be used. The equation and fit to the raw data are given in Figure 2 It is shown in Figure 2 Depicted are examples of a predictive model for cell count versus time in a 10 L pluripotent stem cell bioreactor expansion (right) and sample-based data with model fit (left) (note that the solid data points at 0 and 66 hours were excluded from the model fit).
[0041] To generate predictive viable cell biomass correlations, the Figure 1 and Figure 2 The data are time matched and plotted relative to each other to obtain Figure 3 The data are well described by a quartic model (r2=0.993), but those skilled in the art will appreciate that there are many other models that can be used.
[0042] use Figure 1-Figure 3 The inventors found that Figure 3The equations described in can be used to predict or determine a close approximation of the viable cell density of a PSC bioreactor culture directly from continuous process pH data. Examples of such predictions are given in Figure 4 It is shown in Figure 4 Exemplary continuous process pH data (solid black line) and sample-based cell counts (circles) are depicted. Figure 3 The equations developed and presented in this paper are applied to the generation of pH data for continuous processes. Figure 4 The cell mass prediction is shown in dash-dot lines.
[0043] The above mathematical description also applies to 1) dissolved oxygen when the operation is uncontrolled (i.e., the process value is allowed to decrease as dissolved oxygen is consumed), 2) dissolved oxygen in controlled operation (i.e., dissolved oxygen is held constant and increases in the control value (CV or sensor "call" signal) are used for prediction), 3) oxygen flow from a mass flow controller, and 4) oxygen uptake rate (OUR) calculated from process data. The relationship demonstrated above for pH, combined with the example of dissolved oxygen, results in a two-variable relationship that can be used to predict biomass in a unified multivariable equation or in a parallel redundant manner. The importance of making predictions from more than one continuous process variable is described.
[0044] refer to Figure 5 In another example involving a 10L PSC expansion, the inventors examined dissolved oxygen relationships in a 10L vessel as an example of the advantages of using a multivariate predictive approach. In this example, a PID (proportional, integral, derivative) feedback controller was not set up for proper control and the continuous process data was not amenable to the mathematical processing described above. The process value data ( Figure 5 , top) is confounded by the switch between air and pure oxygen supply (on ~2 days). However, the inventors observed that the control value data ( Figure 5 , bottom) show an increasing trend, which can be additionally exploited into multivariate cell prediction models.
[0045] In summary, the above relationships determined by the inventors have allowed the development of sample-free scale-up seed cohorts. In one embodiment, the predicted biomass values are used to decide when and how much to increase the feed rate provided to a vessel or bioreactor, and are also used to inform decisions to harvest intermediate and final manufacturing stages (between 2-4e6 cells / mL).
[0046] Feed control
[0047] Figure 6Depicted are exemplary data of feed control enabled by biomass modeling, and improvements to existing manufacturing processes enabled by predictive biomass measurements as described above. Initially, from day 2 ( Figure 6 Feed the process at a low rate starting with the feed in 2). Samples are taken daily and the cells will usually expand to the point where increased feed is required by day 4 ( Figure 6 The feed rate was increased by 1 in the middle. This would produce growth kinetics capable of producing approximately 1e6 cells / mL (lower plot line). For example, because the predictive cell yield feedback described herein is based on real-time pH and DO sensor readings, and thus more frequently than physical samples can be collected, the inventors were able to increase the feed rate earlier using the multivariate predictive cell yield determination described herein (middle plot line) and drive the process to higher cell titers (4-5e6 cells / mL) more aggressively (upper plot line).
[0048] Autologous Example – T Cell Manufacturing
[0049] Similar mathematical methods were applied to the manufacture of autologous T cell products in a shaker bioreactor, demonstrating the cell type and platform agnostic nature of the methods described herein. Continuous data (pH and DO) and sample-based data (cell counts) were subjected to curve fitting as previously described and as Figure 7 As shown in Figure 7 Depicted are exemplary model fits for continuous DO data (left), continuous pH data (center), and sample-based cell counts (right). Two unique time-matched fits (Excel-based polynomial and jmp-derived logistic 4 parameter Hill) were generated for each of pH and DO relative to the cells, as well as a multivariate relationship including both pH and DO, and finally the average of all 5 fits.
[0050] As a demonstration of the utility of multivariate and parallel redundant predictive models, Figure 8 Several series shown in show a common theme of increased errors at early time points. This is understandable and attributable to the system because in the early stages of cell expansion, changes in pH and DO are small. However, as cell expansion proceeds and the dynamic range of pH and DO is traversed, the inventors have found that the ability to predict cell biomass yield based on the correlation between cell biomass yield and pH / DO reduction is increased. Figure 8 It was also shown that the averages of multiple replicates (i.e., DO-predicted cells, pH-predicted cells, JMP-4P Hill pH cells, and JMP-4P Hill DO cells), multiple variables, and all correlations represent comparable and predictable modes of operation. Figure 8 As indicated in the legend, Figure 8 Each segment of the bar graph depicts 6 bars. For the JMP-4P Hill pH cell bar shown at 168 hours, although difficult to discern, the data series is shown, where the value is 1%; because the percentage difference from the actual value is quite small, it is overlaid in the figure by the "average of 5 predicted values" bar. Figure 8 It was shown, and the inventors found, that as the prediction progressed over time, the difference from the actual value decreased (ie, the prediction accuracy increased), which demonstrated to the inventors the value of recursive fitting as it adjusted in real time to the data retrieved from the sensor.
[0051] Since growth curves and kinetic data can be generated from continuous process data, Figure 2 and Figure 7 The equations describing growth in can be recursively fitted as real-time pH and DO sensor data are received. The equations in Table 1 below can be updated in real time (as the operator enters a start time (time 0 or "T0" in Table 1) and a target cell density) and presented in a manufacturing graphical user interface (GUI) such as Fig. 9 90 shown in the GUI, Fig. 9 Depicts an example interface (depicts Xuri TM The user interface includes a real-time updated countdown time to harvest 92 and a timestamp 94 for the harvest time. Of particular interest to manufacturers will be the "time to harvest" and "harvest time" to allow for scheduling by the process, operator, and patient. In one embodiment, this may be as shown in Tables 1 and Fig. 9 or transmitted directly to a cloud-based schedule for personnel, downstream resources, and logistics.
[0052]
[0053] Table 1: Predictive analytics calculated by recursive growth curve fitting fed by real-time correlation with pH and DO (user inputs are shaded for this embodiment)
[0054] Described below are various additional embodiments based on the correlation-based predictions discussed above. The present specification generally relates to methods and systems for electronically monitoring two or more bioreactor sensor readings to determine or predict the cell biomass yield in a bioreactor in a non-invasive manner (i.e., in a manner that does not require physical sampling or removal of aliquots of the matrix from the bioreactor) based on a predetermined (e.g., where a model system has been determined) or algorithmically recursively fitted or adjusted correlation between the sensor readings and the cell biomass yield, as shown above. Once the correlation is determined, the approximate cell biomass or bioreactor product amount can be predicted based on the observed sensor readings. The embodiments described herein employ two or more sensor readings, and the predictions can be based on independent, parallel considerations of multiple sensor readings (to reduce the risk of errors from fluctuations of a single sensor) or in a multivariate manner. In an embodiment, as Figure 2 As shown in , outliers in sensor readings can be ignored in the predictive determination of cell biomass or bioreactor product yield.
[0055] In various described embodiments, it is expected that the methods and systems herein facilitate cell biomass prediction without physically removing aliquots of the matrix from the bioreactor, which is expected to reduce or mitigate the risks and costs previously described herein. In various embodiments, it is expected that the methods and systems herein facilitate monitoring and recording bioreactor data and predicted cell biomass yields, which tend to facilitate the scheduling of "harvest time", which in turn facilitates the scheduling of processes, operators and patients, as previously described. The determination of the approximate amount of the product can be 1) performed continuously in real time by the processor when A) sensor data is received (whether the sensor data is received continuously or at discrete intervals), or B) performed on demand when a user input requests to determine the predicted approximate amount of the product based on the sensor data received up to that time point, or 2) performed by the processor based on the continuous reception of sensor data, or based on the receipt of sensor data at discrete time intervals (which may or may not be preconfigured, or at the same interval as the processor is configured to determine the approximate amount of the product), in which case the user request to determine the predicted approximate amount of the product can be based on the most recently received sensor data. In either case, the system can poll updated sensor data upon receiving a user request for determining a predicted approximate amount of a product. The methods and systems described herein can also be used to predict or determine other possible outputs without physically removing or sampling an aliquot of substrate from a bioreactor, such as metabolite output or any other product of a bioreactor cell manufacturing process.
[0056] Although the inventors have demonstrated a correlation between a decrease in pH and DO and a decrease in cell biomass by sampling cell biomass and simultaneous sensor readings of pH and DO, the inventors have used this correlation to predict or determine approximate cell biomass yields without invasive sampling, it should be understood that various other sensor readings of the bioreactor culture medium matrix throughout the cell therapy manufacturing run can be used to determine similar correlations. Exemplary sensor readings (i.e., different physical properties that can be measured by different sensor types) that can be used in the embodiments described herein include, but are not limited to, temperature, conductivity, NO / NOx, volatile organic compounds (VOCs), ozone, chlorine, reduction-oxidation (redox) potential, agitation, inward airflow, outward airflow, gas consumption, capacitance, optical density, particle size, metabolite concentration, pressure, cytokine or growth factor concentration, and container weight. It should also be understood that the methods and systems described herein can be used with any type of bioreactor (including, but not limited to, batch, fed-batch or perfusion / continuous stirred tank reactors and shaker reactors) for growing cells from cell culture based on any suitable culture medium or matrix (including liquid or solid culture medium) and container types other than bioreactors (and therefore, the use of the term "bioreactor" herein is intended to cover all such bioreactors and containers). In addition, while the inventors have demonstrated the applicability of the methods and systems described herein for pluripotent stem cell (PSC) and T cell biomass prediction, the methods and systems described herein can also be used to generally predict the yield of immune cells, stem cells, stem cell-derived cells, and virus-producing cells, as well as other types of cells (including natural killer cells (NK), T regulatory cells (T-reg), tumor infiltrating lymphocytes (TIL), mesenchymal stem cells (MSC), etc.), as well as differentiated cells based on pluripotent stem cells (i.e., PSC-derived cells) and viral vector or exosome-producing cells.
[0057] In other embodiments, recursive model fitting, which can be performed in real time or at discrete time intervals, can be used to link process data from multiple sensor inputs to predictive cell quality, metabolite output, process input (such as feed) requirements or other cascade controls (or some combination thereof). In one aspect, the predicted cell biomass (or other predicted amount, such as metabolite output) can be used as an input to a feedback controller for providing process input (e.g., feed) to a bioreactor to achieve the desired growth trajectory of cell biomass, as described in more detail below. As used herein, "process input" includes any culture medium addition, such as feed, gas flow [w1] (such as, to maintain a certain pH level or change pH according to, for example, cell density and growth rate), dissolved oxygen (DO), specific matrix culture medium components, or any other input that can be provided to a bioreactor (including, for example, agitation of the bioreactor), and also includes any corresponding volume removal with cell retention (i.e., perfusion to remove impurities).
[0058] refer to Fig.10 In one embodiment, a computer-implemented method 1000 for approximating a product yield in a bioreactor includes: providing 1002 a bioreactor containing living cells in a matrix for culturing a product derived from the cells or a product of the cells (i.e., the cells themselves) during a manufacturing process (such as an autologous or allogeneic cell or gene therapy manufacturing process, or, for example, manufacturing of a viral vector) for a period of time; providing 1004 two or more sensors for measuring two or more different conditions or physical properties of the matrix or bioreactor contents (such as those described above) during the period of time of the manufacturing process; receiving 1006 sensor data from the two or more sensors at a processor; and upon receiving the sensor data (including in substantially real time after receiving the sensor data, or not in substantially real time), determining 1008 an approximate amount of the product via the processor based on an algorithm that correlates the sensor data and the product yield (such as described above) (predetermined based on a determined correlation model, or recursively fitted or adjusted to process the data, as described herein, which would not require prior determination of the correlation model) that correlates the sensor data and the product yield (such as described above). Reference Fig.11According to further embodiments, a computer-implemented method 1100 for approximating product yield in a bioreactor includes: providing 1102 a bioreactor containing living cells in a matrix for culturing a product derived from the cells or a product of the cells (i.e., the cells themselves) during a manufacturing process (such as an autologous or allogeneic cell or gene therapy manufacturing process, or, for example, manufacturing of a viral vector) for a period of time; providing 1104 two or more sensors for measuring corresponding two or more different conditions or physical properties (such as those described above) of the matrix or bioreactor contents during the period of time of the manufacturing process; receiving 1106 sensor data from the two or more sensors at a processor; and determining 1108 an approximate amount of the product based on a recursively adjusted or recursive algorithm (determined as described herein) that relates the sensor data to the product yield by considering the sensor data for the two or more different physical properties independently or by multivariate consideration of the sensor data for the two or more different physical properties. As described herein, the inventors have surprisingly discovered the ability to approximate bioreactor product yield based on multiple sensor readings. In particular, this has been demonstrated for cell therapy manufacturing processes where the manufacturing process product includes pluripotent stem cells or T cells.
[0059] In additional aspects of the embodiments described herein are methods and systems for determining the rate at which a feed is supplied to a bioreactor. Fig.12 In one embodiment, method 1200 includes: providing 1202 a target or desired product density or amount; determining 1204 an approximate amount of product via a processor according to methods and systems described herein; determining 1206 whether the determined approximate amount of product is less than the amount required to achieve the target product density or amount within a time period of the manufacturing process (or some other time period, which may be input by a user), and if so, determining 1208 a rate of feed supply to the bioreactor required to achieve the target product density or amount within a time period of the manufacturing process (or some other time period, which may be input by a user); and adjusting 1210 the rate of feed supply to the bioreactor to the determined rate of feed supply. In yet another aspect of the embodiments described herein are methods and systems for determining a supply rate of a process input to a bioreactor. Reference Fig.13In one embodiment, method 1300 includes: providing 1302 a target or desired product density or amount; determining 1304 an approximate amount of a product via a processor according to methods and systems described herein; comparing 1306 the determined approximate amount of the product to the target product density or amount via the processor; determining 1308 a supply rate of a process input to the bioreactor required to achieve the target product density within a time period of the manufacturing process (or some other time period, which may be input by a user) based on the comparison; and adjusting 1310 the supply rate of the process input to the bioreactor to the determined supply rate of the process input. The adjustment may be an increase or decrease in the supply rate of the process input (which may also include increasing or decreasing the rate at which impurities or other materials are removed from the bioreactor while maintaining cell quality, such as by perfusion, as described herein). The adjustment may be a result of a determination based on the comparison 1306 of the determined approximate amount of the product to the target product density or amount that the determined approximate amount of the product is below, at, or above the target product density or amount. In other embodiments, the methods and systems described herein can be used to determine the approximate amount of any process metric associated with a bioreactor manufacturing process, not just the approximate product amount. In yet other embodiments, the methods and systems described herein can be used to determine the supply rate of a process input to a bioreactor based on a comparison to a target for any such process metric, not just a target product density or amount.
[0060] The methods described above can be based on real-time monitoring and processing of multiple bioreactor sensor readings to determine the predicted cell biomass yield in real time, or can be based on monitoring multiple bioreactor sensor readings at discrete time intervals (which can be preconfigured), or based on user input requirements. It should be understood that in order to determine or predict the approximate amount of a product, such as cell biomass, in substantially real time upon user request (or upon receipt of sensor data (at discrete intervals or continuously), although in some embodiments, the determination or prediction of the approximate amount of the product may not be performed in real time upon receipt of sensor data), the methods described above need to be implemented by a computer. As used herein, the term "real time" includes "substantially real time" to account for lags in computer processing, digital read / write operations, and communication transmissions, as will be known to those skilled in the art. In one embodiment, a system 1400 for implementing the methods described herein may include a bioreactor 1402, which includes two or more sensors 1404 (such as for reading pH and DO of a culture medium matrix), a feed input port 1406, and a waste output port 1408, as well as other inputs and outputs known in the art. The bioreactor may include an onboard processor or controller 1410 for reading sensor data and determining the amount of the predicted cell biomass or the product produced based on a predetermined or recursively modeled correlation. Additionally or alternatively, an independent computing device 1450 with a processor 1452 (described further below) may be used for this purpose. The processor 1410 and / or 1452 (collectively referred to herein as "processor" or "the processor") may determine in real time (or non-real time) and recursively determine based on continuous input from the sensor. In other embodiments, the bioreactor may include (or may have a separate) feedback controller 1412, which is communicatively coupled to one or more feed input ports (or output ports) of the processor and the bioreactor, and based on the predicted cell biomass or other determined predictions, the feedback controller 1412 provides feed to the bioreactor (or removes material from the bioreactor), and the feedback controller may receive and provide feedback control of the bioreactor input (or output) based on the real-time (or non-real-time) prediction of the cell biomass or other determined predictions. In other embodiments, the processor and feedback controller / processor can be contained in the same physical unit, or can constitute the same unit. In other embodiments, any controller or processor for described method and system can be located away from bioreactor. The communication between controller and bioreactor can be connected via wired or wireless, using transmitter, receiver and / or transceiver. Other aspects of the system for performing the method herein are described below.
[0061] refer to Fig.14In addition to the bioreactors and sensors described herein, the system 1400 for performing the methods described herein may also include a computing device 1450, which includes a processor 1452, a memory 1454, an input device 1456, and a communication module 1458. The computing device may also include a receiver, a transmitter, and / or a transceiver 1460. The processor may be communicatively coupled to the memory, the input device, the communication module, and / or any receiver, transmitter, and / or transceiver, and the computing device may be located remotely from the bioreactor or co-located with the bioreactor and communicatively coupled to the sensor as described herein. All processors described herein may be configured to perform any or all method steps suitable for computer implementation. For example, in one embodiment, the processor can be configured to: receive 1006, 1106 (e.g., via receiver 1460 or communication module 1458) sensor data from two or more sensors 1404; and determine 1008, 1108 approximate amounts of products by considering sensor data for two or more different physical properties independently or by considering sensor data for two or more different physical properties in a multivariate manner based on a predetermined or recursively adjusted or recursive algorithm stored in memory 1454 that associates sensor data for two or more different physical properties or conditions (such as those described above) with product yields.
[0062] In embodiments involving determining 1208, 1308 and / or adjusting 1210, 1310 a process input (e.g., feed) supply rate to a bioreactor, the system 1400 may further include a feedback controller 1412 for performing the adjustment of the process input (e.g., feed) supply rate, as discussed above. The feedback controller may be communicatively coupled to the computing device 1450 and the bioreactor 1402, such as coupled to a feed input port 1406 (or output port 1408) of the bioreactor. Upon receiving the determined adjusted process input (e.g., feed) supply rate as determined by the processor according to the methods described herein from the communication module 1458 of the computing device 1450, the feedback controller 1412 may adjust the supply rate of the process input (e.g., feed) to the bioreactor to the supply rate of the process input (e.g., feed) determined by the processor, such as by opening, closing, or otherwise adjusting the flow of the process input (e.g., feed) through the input port 1406 or output port 1408 of the bioreactor.
[0063] The processor 1410 and / or 1452 of the described system 1400 for processing information associated with the process or method described herein can be, for example, any computer processor known in the art that is capable of performing calculations and directing the functions of input, output, calculation and display of data interpreted and / or executed according to the disclosed method. The processor may include any type of processing unit, such as a typical computer processor, controller, microcontroller, microprocessor and / or programmable logic controller (PLC). The information to be processed by the processor may include, for example, information contained in an analog signal or a digital signal and / or a converted signal and / or information contained in a data memory. The processing of information may include, for example, performing calculations on the received signal, such as, but not limited to, vector analysis, picture identification, pattern recognition, frequency analysis / Fourier transform, numerical calculation, machine learning, or, as described herein, applying a predetermined or recursively fitted correlation algorithm to the received sensor data. In some embodiments, the system includes more than one processor, and reference to "processor" herein includes reference to multiple processors, and vice versa.
[0064] In one embodiment, the processor 1410 and / or 1452 communicates with two or more sensors 1404 of the bioreactor 1402. In another embodiment, the processor 1410 and / or 1452 may also communicate with a data storage 1462 and optionally a display 1464 (such as for displaying the GUI 90). Components of the system 1400, such as the bioreactor 1402, the sensors 1404, the computing device 1450, the processor 1410 and / or 1452, the feedback controller 1412, the data storage 1462 and / or the display 1464, and any other components of the system 1400, the bioreactor 1402, or the computing device 1450, may communicate using any electronic wired or wireless means or protocols known in the art for communication, including but not limited to Ethernet. TM 、Bluetooth TM , WiFi TM , infrared, near field communication (NFC), radio frequency identification (RFID), WiMAX TM(fixed or mobile), cellular communication protocols such as GSM, EDGE, GPRS, CDMA, UMTS, LTE, LTE-A, IMS, and any other cellular communication protocol, including but not limited to, for example, up to and including 5G protocols as established under 3GPP, and any other communication protocol suitable for the methods and systems described herein, including any proprietary protocols. The components of the system may exist on the same network or on separate networks, and the network may include any type of network suitable for the systems and methods described herein, including but not limited to wired or wireless personal area networks (PANs), local area networks (LANs), mesh or ad hoc networks, wide area networks (WANs), metropolitan area networks (MANs), virtual private networks (VPNs), and any other suitable network types, and any suitable network configurations or topologies (e.g., token rings, star, bus, mesh, tree, etc.). The systems described herein also include any components required to implement the communications and / or network types employed, such as wireless or wired routers and access points.
[0065] The methods and systems described herein can be implemented on a secure network, access to which can be limited to authorized users by any known means, and the secure network can be protected by known security measures, such as by using a firewall. In some embodiments, authentication may be required before granting access to authorized users, such as in the case of autologous cell manufacturing and / or patient data, and / or in the case of requiring compliance with government regulations (such as Chapter 21 of the U.S. Code of Federal Regulations). Such authentication can be implemented for any one or more system components, such as access to a computing device or machine that accommodates a processor, access to a data storage device, access to a database of a data storage device, access to any sensor or sensor reading of a bioreactor, access to a graphical user interface (GUI) of the system, access to a bioreactor, etc. The security of the methods and systems described herein can also be provided by encrypting communications between system components by any means or protocols known to those skilled in the art, for example, to reduce the possibility of tampering with or destroying pre-programmed related algorithms, such as Internet Protocol Security (IPSec), Transport Layer Security (TLS), Secure Sockets Layer (SSL), etc.
[0066] The systems described herein may also include a data storage 1462 for storing information associated with the described methods. The data storage may include, for example, various types of local storage devices (such as Fig.14) or remote storage devices, such as hard disks or hard drives (of any type, including electromechanical disks and solid-state disks), memory chips, including, for example, random access memory (RAM) and / or read-only memory (ROM), flash memory, optical storage, such as CDs and DVDs, floppy disks, and any other form of optical, physical, electronic and / or magnetic storage device in which or on which information can be stored. The data storage device may include non-volatile memory. In some embodiments, the data storage device may be accessed only via secure data transmission, which can be implemented using one or more known server platforms and security protocols. The information to be stored in the data storage device may include, for example, one or more predetermined algorithms for associating sensor inputs with various outputs such as cell biomass yield; and such predicted determined records, together with associated actions (such as feeding) and / or associated timestamps, unique identifiers, authorized users and associated authentication information for authenticating the user's authorized access to the data storage device or any other system component, and any other relevant information. In operation, the data storage device communicates with the processor.
[0067] The system may also include a display 1464 (eg, Fig.14 , which can be co-located with a processor, e.g., where the processor and the display are part of a computer or server for performing the method steps described herein), for visually presenting information associated with the described methods. The display may include, for example, a computer monitor (e.g., an LCD, CRT monitor, a projection (e.g., a head-up display (HUD) laser), etc. In some embodiments, the visual display may include, for example, a display of a mobile device, such as a tablet computer, a cellular phone, a smart phone, a personal digital assistant (PDA), a personal computer (PC), a laptop computer, an augmented reality display (e.g., Google TM Glass TM or Microsoft TM HoloLens TM ) etc. The information presented on the display may be, for example, Fig. 9 , and / or may include any other information collected during the execution of the methods described herein, prompts for information input associated with one or more steps of the described methods, and / or any predetermined formulas or algorithms, as previously described. The display may also be capable of receiving input (such as, for example, where the display includes a touch screen and is capable of receiving touch input and thus transmitting information to the processor).
[0068] It should be understood that the components of the described system, such as a computer or machine housing a processor, include components known in the art required for its operation, such as a power supply, a network interface (such as a network interface card), a network connection component (e.g., a modem, an Ethernet card, a USB interface card, a FDDI card, a WLAN card, etc.), a receiver, a transmitter, a local memory such as RAM, ROM, flash memory, cache or buffer memory, and / or other types of memory as previously described, a processing unit that can communicate with input / output (I / O) devices, and all required circuits, including buses. As will be known to those skilled in the art, the bioreactor sensor 1404 may also include components specific to the type of sensor. The computing device 1450 such as a computer or machine housing a processor may also include, for example, memory (e.g., hard disk memory, RAM, ROM, flash memory, cache or buffer memory, and / or other types of memory as previously described), connected input devices (e.g., a mouse, keyboard, microphone, etc.), connected output devices (e.g., a display monitor), and local memory for the processor (e.g., registers, cached RAM, such as L1 cache, L2 cache, etc.). Depending on the system components, other components may also be present (eg, a device for communicating with a cellular network may include an antenna, etc.), and it should be understood that such components will be known to those skilled in the art.
[0069] The described system may include one or more redundant sensors for each of two or more sensor types in a bioreactor to provide additional data for determining correlations and thereby further improve the accuracy of predictions of cell biomass yield based on sensor readings (i.e., while each of the two or more sensor types in the methods and systems described herein is a different type of sensor, each different type of sensor may have an associated redundant sensor of the same type). Redundant sensors for each of the two or more sensor types in a bioreactor may also be used in larger bioreactors to ensure that the system is homogeneous. In the event that, for example, sensor information is consistent between redundant sensors, it may be stored in a data storage device (e.g., in a database thereof), and in the event that the captured information is inconsistent between redundant sensors, the corresponding sensor readings may be averaged or ignored altogether, or one of the redundant sensor readings, if not generally consistent with the trajectory of other received sensor data, may be discarded as an outlier, as previously described. The redundancy of the data storage device and / or its database is expected to help ensure the persistence of the stored data and reduce the risk of data loss.
[0070] In some embodiments, the methods provided herein can be implemented using, for example, computer-readable and executable instructions as described above. Thus, another aspect provided herein is a tangible, non-transitory computer-readable medium (i.e., a medium that includes more than just the temporary propagation signal itself), such as memory 1454, including or having stored thereon or therein computer-executable instructions associated with the methods described herein, such as a local or remote hard disk or hard drive (of any type, including electromechanical disks and solid-state disks), a memory chip, including, for example, random access memory (RAM) and / or read-only memory (ROM), a cache, a buffer, a flash memory, an optical memory such as a CD and a DVD, a floppy disk, and any other form of storage medium, in which information can be stored in a volatile or non-volatile manner or thereon, for any duration, including permanent storage or temporary storage. Such computer-executable instructions, if executed by a processor (e.g., a processor of a computing device 1450, such as a computer that houses a processor described herein), cause the processor and / or computing device, computer, or machine to perform any method steps described herein that are suitable for computer implementation. Different implementations of the disclosed methods may include performing some or all of the steps described herein in different orders, or performing some or all of the steps substantially in parallel. The functions or method steps may be implemented in various programming languages, and such code or computer readable or executable instructions may be stored or adapted to be stored in one or more machine readable media, such as described above, which may be accessed by a processor-based system to execute the stored code or computer readable or executable instructions.
[0071] It should be understood that any method steps, modules, components, or systems described herein that are suitable for computer implementation or required for computer implementation in order to achieve real-time or substantially real-time execution may be implemented using computer-readable / executable instructions or operations, which may be stored or otherwise maintained by a computer-readable medium, as described herein.
[0072] Although certain specific embodiments have been referenced herein, various modifications thereof will be apparent to those skilled in the art without departing from the purpose and scope of the present invention as outlined in the appended claims. Any examples provided herein are included solely for the purpose of illustrating the present invention and are not intended to limit the present invention in any way. Any drawings provided herein are merely for the purpose of illustrating various aspects of the present invention and are not intended to be drawn to scale or to limit the present invention in any way. The disclosures of all prior art cited herein are incorporated herein by reference as if set forth in their entirety.
Claims
1. A computer-implemented method for approximating product yield in a bioreactor, the method comprising: providing a bioreactor containing living cells in a matrix, the bioreactor being used to culture a product derived from the cells or a product of the cells during a manufacturing process for a period of time; providing two or more sensors for measuring corresponding two or more different physical properties of said substrate during said time period; receiving, at a processor, sensor data from the two or more sensors; and An approximate amount of the product is determined via the processor by independently considering the sensor data for the two or more different physical properties based on a recursive algorithm that relates the sensor data to the product yield, wherein the recursive algorithm performs a recursive fit as it adjusts the sensor data for the two or more different physical properties in real time.
2. The computer-implemented method of claim 1, wherein the manufacturing process comprises an autologous cell or gene therapy manufacturing process or an allogeneic cell or gene therapy manufacturing process.
3. The computer-implemented method of claim 1, wherein the product comprises an immune cell, a stem cell, a stem cell-derived cell, a virus-producing cell, a natural killer cell (NK), a T regulatory cell (T-reg), a T cell, a tumor-infiltrating lymphocyte (TIL), a mesenchymal stem cell (MSC), a pluripotent stem cell (PSC), a PSC-derived cell, a viral vector-producing cell, or an exosome-producing cell.
4. The computer-implemented method of claim 1, wherein the product comprises a pluripotent stem cell (PSC), a PSC-derived cell, or a T cell.
5. The computer-implemented method of claim 1, wherein the two or more different physical properties include temperature, conductivity, NO / NOx, volatile organic compounds (VOCs), ozone, chlorine, reduction-oxidation (redox) potential, agitation, inward airflow, outward airflow, pressure, container weight, pH, metabolite concentration, cytokine or growth factor concentration, capacitance, optical density, or dissolved oxygen (DO).
6. The computer-implemented method of claim 5, wherein a first of the two or more sensors comprises a sensor configured to measure the pH, and a second of the two or more sensors comprises a sensor configured to measure the dissolved oxygen (DO).
7. The computer-implemented method of claim 1 , wherein: The determining of the approximate amount of the product is continuously performed by the processor in substantially real time as the sensor data is continuously received.
8. The computer-implemented method of claim 1 , wherein: The determining of the approximate amount of the product is based on continuous reception of the sensor data or reception of the sensor data at discrete time intervals and is performed by the processor at the discrete time intervals.
9. The computer-implemented method of claim 1, wherein the processor is configured to ignore outliers in the sensor data when determining the approximate amount of the product.
10. A method of determining a supply rate of a process input to the bioreactor of claim 1, the method comprising: Providing target product density; Determining the approximate amount of the product by the method of claim 1; comparing, via the processor, the approximate amount of the product to the target product density; determining, via the processor, the supply rate of the process input to the bioreactor required to achieve the target product density within the time period based on the comparison; and The supply rate of the process input to the bioreactor is adjusted to the determined supply rate of the process input.
11. A system for approximating product yield in a bioreactor, the system comprising: a bioreactor for culturing a product derived from or in a matrix of living cells during a manufacturing process for a period of time; two or more sensors for measuring corresponding two or more different physical properties of the substrate during the time period; a computing device communicatively coupled to the two or more sensors, the computing device comprising: processor; Memory; and Communication module; the processor being communicatively coupled to the memory and the communication module; The processor is configured to receiving sensor data from the two or more sensors; and Based on a recursive algorithm stored in the memory that relates the sensor data to the product yield, an approximate amount of the product is determined by independently considering the sensor data for the two or more different physical properties, wherein the recursive algorithm performs a recursive fit as it adjusts the sensor data for the two or more different physical properties in real time.
12. The system of claim 11, wherein the product comprises an immune cell, a stem cell, a stem cell-derived cell, a virus-producing cell, a natural killer cell (NK), a T regulatory cell (T-reg), a T cell, a tumor-infiltrating lymphocyte (TIL), a mesenchymal stem cell (MSC), a pluripotent stem cell (PSC), a PSC-derived cell, a viral vector-producing cell, or an exosome-producing cell.
13. The system of claim 11, wherein the product comprises a pluripotent stem cell (PSC), a PSC-derived cell, or a T cell.
14. The system of claim 11, wherein the two or more different physical properties include temperature, conductivity, NO / NOx, volatile organic compounds (VOCs), ozone, chlorine, reduction-oxidation (redox) potential, agitation, inward airflow, outward airflow, pressure, container weight, pH, metabolite concentration, cytokine or growth factor concentration, capacitance, optical density, or dissolved oxygen (DO).
15. The system of claim 14, wherein a first of the two or more sensors comprises a sensor configured to measure the pH, and a second of the two or more sensors comprises a sensor configured to measure the dissolved oxygen (DO).
16. The system according to claim 11, wherein: The determining of the approximate amount of the product is continuously performed by the processor in substantially real time as the sensor data is continuously received; or Based on the continuous reception of the sensor data or the reception of the sensor data at discrete time intervals, the processing is executed by the processor at the discrete time intervals.
17. The system of claim 11, wherein the processor is configured to ignore outliers in the sensor data when determining the approximate amount of the product.
18. A system for determining a supply rate of a process input to the bioreactor of claim 11, the system comprising: The system of claim 11, for determining an approximate amount of said product; and a feedback controller communicatively coupled to the computing device of claim 11 and the bioreactor; The processor is further configured to comparing the approximate amount of the product to a target product density; determining the supply rate of the process input to the bioreactor required to achieve the target product density within the time period based on the comparison; and transmitting, via the communication module, to the feedback controller an adjusted supply rate of the process input to the bioreactor; wherein, upon receiving the adjusted supply rate of the process input from the communication module of the computing device, the feedback controller adjusts the supply rate of the process input to the bioreactor to the supply rate of the process input determined by the processor.
19. A non-transitory computer readable medium comprising computer executable instructions for approximating a product yield in a bioreactor for culturing a product derived from or in a matrix during a manufacturing process for a period of time, the bioreactor comprising two or more sensors for measuring respective two or more different physical properties of the matrix during the period of time, the computer executable instructions when executed causing a processor to: Based on a recursive algorithm correlating sensor data received from the two or more sensors and the product yield, an approximate amount of the product is determined by independently considering the sensor data for the two or more different physical properties, wherein the recursive algorithm performs a recursive fit as it adjusts the sensor data for the two or more different physical properties in real time.
