Method and system for optimizing culture conditions in a culture process - Patents.com

Abstract

A method and system for iteratively improving culture conditions for at least one culture output in a culture process is provided. The method includes: (a) providing a continuous culture of cells in a bioreactor, where the cells are cultured under culture conditions including a set of culture parameters, where the culture of cells produces at least one culture output in a first measurement; (b) while maintaining the cells in continuous culture, iteratively testing different culture conditions by: (i) selecting new culture conditions including a new set of culture parameters that are predicted to improve the measurement of the culture output; (ii) perturbing the cells with the new culture conditions; and (iii) determining a new measurement of the culture output, where the testing is repeated to produce multiple improvements in the measurement of the culture output. The iterative process can be performed by automated artificial intelligence. TIFF2025503486000002.tif62168

Description

[Technical field]

[0001] Statement Regarding Federally Sponsored Research none.

[0002] REFERENCE TO RELATED APPLICATIONS This application claims the benefit of the filing date of U.S. Provisional Application No. 63 / 292,397, filed December 21, 2021, the contents of which are incorporated by reference in their entirety herein.

[0003] Names of the parties to the joint research agreement none.

[0004] Sequence Listing none. [Background technology]

[0005] background The burgeoning biomanufacturing and synthetic biology industries rely on bioreactors (also called fermenters) for the bioconversion of raw materials into value-added products. Biomanufacturing is an emerging field that can be leveraged to produce a large portion of the physical materials in the world economy, not only for improved performance but also with a better environmental footprint to combat climate change. Recent estimates put the total annual economic impact of synthetic biology at $2-4 trillion by 2040. Two-thirds of this impact will be in non-healthcare sectors, namely agriculture, consumer products, biomaterials, biochemicals, bioenergy, and many others. Of all the potential applications of synthetic biology, biomanufacturing of alternative proteins, biomaterials, biochemicals, and biofuels will experience the greatest growth over the next 10-20 years.

[0006] Biomanufacturing may be one of the most technically complex industries, yet its underlying production process, batch fermentation, has not changed in decades. Batch reactors are slow, repetitive, and continuous in nature. Bioreactors are very expensive to acquire, in addition to having long lead times. Due to the prohibitive price and limited availability of bioreactors, "tank-time" is the biggest bottleneck for optimizing fermentation processes (bioprocesses). For example, using a Design of Experiment (DoE) approach to identify optimal process conditions may require hundreds of bioreactor runs, each with a different set of operating conditions. Assuming that each run takes an average of one week, it would take two full years to achieve 100 runs with a single bioreactor. Even established companies with multiple bioreactors have a limited amount of "tank-time" they can spend per project, which may force them to reduce the number of conditions they test. Many projects are "shelved" during bioprocess scale-up due to lack of improvement over time. Similar limitations exist in bioreactor strain screening, where limited tank time often results in only one "baseline" process condition being used to screen all strains, resulting in severe underperformance of the true optimum strains due to suboptimal bioprocess conditions. Tank time limitations due to high bioreactor costs lead to poor process optimization and high rates of false negative results in strain screening, resulting in missed economic and financial opportunities and suboptimal performance at large scale. Summary of the Invention

[0007] overview Disclosed herein is a methodology for generating high-throughput fermentation process data for rapid process optimization and strain screening for multiple process parameters. The high-throughput fermentation platform leverages autonomous agents, such as machine learning (ML)-enhanced real-time optimization based on specific resource consumption rate (mass of resources consumed per cell per hour), specific production rate (mass of product produced per cell per hour), and production economics. This real-time automated optimization uses a robust continuous fermentation platform. Replacing traditional fed-batch bioreactor systems with continuous bioreactors minimizes equipment downtime, increases volumetric productivity, and reduces operational costs. Unlike batch fermentation, which tests only a single condition per tank run, continuous bioreactors allow for temporal multiplexing of testing process conditions. Any number of conditions can be tested during a run as needed. This results in a 5-20x reduction in process optimization time and a reduction in the number of tanks required to complete an optimization campaign. The disclosed technology provides significant productivity gains as measured by data generated per tank per day, addressing a major pain point in biomanufacturing and eliminating a bottleneck in synthetic biology optimization. Successful realization of continuous optimization anticipates future advances in high-throughput automation technologies and continuous downstream processing.

[0008] The first step of this high-throughput method is to determine baseline strain stability in a continuous bioreactor (e.g., turbidostat) and baseline productivity in a (fed-)batch reactor. Continuous operation tests strain stability and provides the number of generations that can be performed in a continuous system before productivity declines; this determines the range of time that the continuous reactor can be operated. For (fed-)batch reactor operation, initial growth rate, productivity, and yield are determined prior to optimization. Strains exhibiting constitutive or inducible expression of the production pathway are cultivated in a turbidostat at constant biomass concentration and fermentation input parameters. For inducible processes, inducer / inhibitor molecules are added / removed after initial growth to induce the production pathway. After 24 hours, the turbidostat system reaches a steady state where constant product titers and specific productivity are observed. Strain breakage causes a gradual decline in titer and productivity over time, which is observed and measured by offline sample analysis. Strain stability is assessed based on the time elapsed until strain breakage. Since unstable strains are not suitable for commercial production, this step provides a quick go / no-go decision, allowing time and money to be saved before further fruitless development is carried out.

[0009] The second step is to optimize the best culture conditions for the production organisms, since a robust growth environment is the basis for production. The growth rate can be optimized using a turbidostat. Strains exhibiting constitutive or inducible expression of production pathways can be cultivated in a turbidostat where the biomass concentration is constant but the fermentation input parameters are varied. After 24 hours, the turbidostat system reaches a steady state where all nutrients are in excess and the production organism reaches its maximum specific growth rate. Different input fermentation parameters are applied to perturb the system from the current steady state to a new steady state. This new steady state has a new maximum specific growth rate that is different from the previous specific growth rate. This process continues until a maximum specific growth rate is identified or the generation time defined in step (1) is reached. To perturb the system, the input fermentation parameters are implemented. The inputs can be changed stepwise or sinusoidally rather than allowed to reach a steady state. The two conditions can be ranked by comparing the rate or trend of change of the specific growth rate between the two sets of input fermentation parameters. Process conditions to be tested are randomized or repeated to increase the statistical power of the results. For certain common production hosts such as Escherichia coli, Bacillus species, Pichia pastoris, and Saccharomyces cerevisiae, this step may be optional since the growth ranges on most common production platforms are already known.

[0010] The third step is to optimize the production organism for the best production conditions. This approach not only scans an array of fermentation conditions for the highest productivity and yield, but also presents a more comprehensive way of screening strains since it is no longer limited to a single screening condition as in fed-batch mode. To optimize for specific productivity and yield, a chemostat can be used. For inducible strains, inducer / inhibitor molecules are added / removed after initial growth to induce the production pathway. After 24 hours, the chemostat system reaches a steady state where all nutrients except one are in excess and the production organism reaches a steady state. At the steady state, the rate of addition of this rate-limiting nutrient will determine the growth rate of the system. Different input fermentation parameters are applied to perturb the system from the current steady state (state 1) to a new steady state (state 2-N). This new steady state will result in a new maximum specific productivity and yield that is different from the previous specific productivity and yield. This process continues until maximum specific productivity and yield are observed or until the generation time to strain stability is reached. When input fermentation parameters are implemented to perturb the system, the inputs can be varied stepwise or sinusoidally. The process conditions tested are randomized or repeated to increase the statistical power of the results. Comparing the rate or trend of change in specific growth rate between two sets of input fermentation parameters allows the ranking of the two conditions.

[0011] Data generation in high-throughput fermentation can be automated by computational algorithms, such as machine learning-based algorithms. For example, reinforcement learning, one of the three basic machine learning paradigms, can be used to take actions during a continuous turbidostat or chemostat process to maximize desired fermentation output parameters. Other mathematical optimization methods, such as dynamic programming, gradient descent, and simplex algorithms, can also be applied. This machine learning-assisted approach would dramatically reduce the conditions that need to be tested and would also identify optima that may be counter to human intuition.

[0012] The fourth step is to validate the optimized parameters in a fed-batch bioreactor. The newly optimized fermentation input parameters, such as pH, temperature, dissolved oxygen, carbon feed rate, nitrogen feed rate, and other chemical compositions of the medium, will be tested to verify and validate the optimization pipeline that has led to significant improvements in productivity and / or production economics. [Brief description of the drawings]

[0013] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments and, together with the description, serve to enable one of ordinary skill in the relevant art to utilize these and other embodiments which will be apparent to one of ordinary skill in the art. The present invention will be more particularly described in conjunction with the following drawings: [Figure 1] FIG. 1 shows an exemplary overall process for optimizing culture conditions for cell growth and product production by cells cultured in a bioreactor. [Diagram 2] Figure 2 shows the three process steps in more detail. [Diagram 3] FIG. 3 shows a more detailed exemplary version of the process for optimizing culture conditions. [Figure 4] Figure 4 shows a schematic diagram of reinforcement learning. [Diagram 5] Figure 5 shows results from a turbidostat where temperature and pH were held constant over time and cells were maintained at a constant concentration (AU). [Figure 6] A comparison of the optimization process performed in continuous culture with that performed in batch culture is shown in Figure 6. In continuous culture, multiple different culture conditions can be tested during one run, whereas under batch conditions, each run is characterized by a single culture condition. [Figure 7] FIG. 7 shows the progress of an optimization experiment, where an intelligent agent selects new culture conditions that are predicted to improve the culture output; in this case, the prediction is made based on the prior result. [Figure 8] 8A and 8B show an exemplary bioreactor. [Figure 9] FIG. 9 illustrates an exemplary computer system. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0014] Detailed Description I. Introduction Provided herein is a method for optimizing culture conditions for a desired culture output. In contrast to conventional methods that rely on (fed)batch culture methods, the optimization method described herein is performed in a continuous culture.

[0015] A. Continuous culture The main difference between batch and continuous culture is that batch culture is a closed system in which cell culture, such as fermentation, is carried out with a constant amount of nutrients, whereas continuous culture is an open system in which nutrients are added continuously to the culture. Batch culture is characterized by an initial rapid cell growth followed by a plateau after nutrients are exhausted and cell density inhibits further growth. Continuous culture is characterized by a constant growth rate and / or a constant rate of culture output, as nutrients are continuously added to the system and cell density is maintained by the withdrawal of excess cells from the culture.

[0016] Cultivation processes include any culture of cells to produce a culture output. Such processes include, but are not limited to, processes referred to as "fermentation" and "cell culture." "Fermentation" generally refers to the enzymatic conversion of one molecule to a different molecular species (e.g., sugar to ethanol), usually by a single-celled organism. "Cell culture" generally refers to the culture of a single cell, e.g., a mammalian cell or an insect cell, to produce a product, such as a polypeptide or a value-added organic compound. These processes further include synthetic biology methods, in which cells are genetically engineered to produce a molecular output, e.g., by intracellular expression of enzymes along a synthetic pathway.

[0017] B. Bioreactor Cultivation of cells is typically carried out in a bioreactor. A bioreactor is a container or vessel configured to contain a culture medium and includes components for maintaining or altering various culture parameters. This includes, for example, a temperature regulator and thermometer for measuring temperature, a source of nutrients, and sensors, such as a pH meter, for measuring the culture parameters. Bioreactors can also include stirring devices, such as paddles, and ports for inputting nutrients and other chemicals and for outputting the culture medium (which may contain cells). Bioreactors can include feeds for adding nutrients and other molecules, as well as inlets for withdrawing the culture medium from the bioreactor.

[0018] Two types of bioreactors used for continuous culture are the turbidostat and the chemostat. In both cases, nutrients and other chemicals are added to the cell culture at a rate called the dilution rate. The dilution rate is determined by feedback from sensors in the culture to maintain desired parameters.

[0019] A turbidostat is a continuous bioreactor in which the cell density (e.g., measured by turbidity) is kept constant. Additionally or alternatively, the turbidostat culture volume can be kept constant. The cell density or biomass is determined by proxy measures such as turbidity or capacitance. Once the cells reach a steady state, the growth rate will be constant if the cell density is maintained. In this system, the maximum or desired growth rate can be achieved by maintaining nutrients in excess. The maximum growth rate varies depending on the medium formulation and process conditions. In certain embodiments, while culturing a cell culture under turbidostat mode, the growth rate of the culture is maximized and / or set to a desired rate, resulting in limited bioproduction of the culture product. In certain embodiments, while culturing a cell culture under turbidostat mode, no culture product is produced. In certain embodiments, the turbidostat is coupled to a volume sensor to maintain both turbidity and volume constant by withdrawing culture from the turbidostat when the volume exceeds a set point.

[0020] A chemostat is a type of continuous bioreactor that maintains steady state culture conditions and therefore growth rate. Culture conditions are maintained by continuously adding medium to the culture to maintain the concentrations of nutrients and other chemicals, while simultaneously removing medium to keep the culture volume constant. Growth rate can be physically controlled by varying the rate of addition of chemicals to the chemostat and by varying the rate of medium removal.

[0021] Figures 8A and 8B show an exemplary bioreactor. Figure 8B shows a bioreactor connected to a computer with a user interface that shows the parameters of the culture, rather than receiving user instructions to control the culture conditions. Bottles containing different types of nutrients are connected to the culture via tubes. The bottles can also receive excess culture medium from the culture.

[0022] Chemostats and turbidostats generally have volumes of about 500 mL to 50 L. This includes, for example, volumes of about 1 L to about 10 L. Industrial size fermentors can be as large as about 5 liters to 500 liters.

[0023] Bioreactors are commercially available from, for example, Sartorius (Goettingen, Germany) and ThermoFisher (Waltham, Mass.).

[0024] C. Culture conditions and parameters Culture conditions include a number of culture parameters for culturing cells, referred to herein as a "parameter set." A change in the measured value of any culture parameter in the parameter set creates a new culture condition. Culture parameters include, but are not limited to, temperature, pH, dissolved oxygen, carbon feed rate, and nitrogen feed rate. Other culture parameters include, for example, nutrient concentrations. Compounds (whose concentrations can be manipulated) include, for example, metals (e.g., iron, zinc, cobalt, copper, nickel, manganese, molybdate, selenite, and other transition metals), vitamins (e.g., niacin, pyridoxine, riboflavin, pantothenate, aminobenzoates, thiamine, biotin, cyanocobalamin, folic acid), inducers, salts, nitrogen sparing rate, aeration rate, oxygen sparge rate, carbon dioxide sparing rate, phosphate, sulfate, chloride, acetate, citrate, and other anion salts, magnesium, calcium, Sodium, potassium, ammonium and other cationic salts, boric acid, choline, ascorbic acid, lipoic acid, nicotinic acid, inositol and other vitamins, antifoaming agents (e.g., Antifoam 204, Antifoam A, Antifoam C), amino acids (e.g., glutamic acid, leucine, and tryptophan), nucleic acid bases (e.g., adenine, cytosine, thymine, uracil, and guanine), complex nutrients (e.g., yeast extract, peptone, tryptone, and casamino acids), macronutrients, and / or micronutrients. For cell culture of eukaryotic cells, such as mammalian cells, the culture parameters can include any of the following: concentration of cell growth factors, CO2 concentration, agitation speed, and antibiotic concentration.

[0025] The carbon source can be a sugar (e.g., glucose, xylose, sucrose, glycerol, or acetate), molasses, malt extract, starch, dextrin, pulp, CO or CO. The nitrogen source can be, for example, an amino acid or polypeptide, urea, an ammonium salt (e.g., ammonium sulfate, ammonium phosphate, or ammonia), corn steep liquor, yeast extract, peptone, and soybean meal.

[0026] D. Culture output The culture output can be any measurable characteristic of the culture to be optimized, including, for example, cellular biomass, molecular products, and product production or proxies for cell health based on cell physiology.

[0027] In certain aspects, the culture output is the biomass of the cells themselves. In this case, the culture output to be optimized may be the cell growth rate. The cell growth rate may be measured as a function of the change in turbidity over time.

[0028] In other aspects, the culture output is a chemical product produced by the cells ("culture product"). Culture products can be molecular entities that are products of fermentation or gene expression, and are typically products that are harvested from the culture and commercialized. Culture products contemplated herein include polypeptides, e.g., proteins, enzymes, antibodies. Culture products also include organic molecules that are products of synthetic pathways within the cells, e.g., via enzymes. Such products include, for example, industrial chemicals. These include, but are not limited to, flavors (e.g., vanillin); fragrances (e.g., aldehydes, coumarins, indoles), amino acids, organic acids (e.g., citric acid, lactic acid, acetic acid); alcohols (e.g., ketones such as ethanol, isopropanol, acetone); fatty acids (e.g., palmitic acid, oleic acid). Culture output can be measured directly as a concentration or as a function of amount, e.g., volumetric production rate, product titer, product yield, or specific production rate.

[0029] In other embodiments, the culture output is a surrogate for output production ("culture proxy"), for example based on cell physiology. A culture proxy is a measurable parameter that indicates the health and / or physiology of a culture of cells. These include, but are not limited to, specific CO2 production rate, specific O2 consumption rate, organic acid profile, metabolite profile, by-product profile, production economics (cost to produce 1 g of product under specific cell culture conditions). Specific rate can be defined as the rate of change of a compound per bacterium per hour, and volumetric production rate can be defined as the rate of change of a compound per liter of fermentation medium per hour.

[0030] E. Cell Culture The cells cultured in the methods disclosed herein can be any type of cell. This includes prokaryotic and eukaryotic cells. Prokaryotic cells can include bacteria and archaea. Eukaryotic cells can include fungi, plant cells, and animal cells. Fungi include, for example, yeast. Plants include, for example, algae. Animal cells include cells from any phylum, including arthropods (e.g., insects, shrimp, lobsters, crayfish, and crabs), and chordates (e.g., fish, amphibians, reptiles, birds (e.g., chickens or turkeys), and mammals (e.g., humans or non-human animals, e.g., cows, rams, goats, pigs, horses, dogs, cats, primates)).

[0031] Cell lines include, for example, CHO (Chinese Hamster Ovary cells), BHK21 (Baby Hamster Kidney), NS0, Sp2 / 0 mouse cell lines, insect cells (e.g., SP9, Sf9, sf21, S2), tobacco BY-2 cells, rice (Oryza Sativa) and algal cells.

[0032] Cells that produce molecular products may do so constitutively or upon induction. In the case of inducible systems, an inducer, such as a molecule or light if the promoter is light-inducible, may be introduced into the culture. In the case of stress-inducible promoters, a stressor is introduced. For example, in the case of phosphate starvation-inducible promoters, phosphate may be limited or removed from the culture.

[0033] F. Measurements Measurements of variables such as culture output can be any combination of numbers and words. Measurements can be any scale, such as nominal (e.g., names or categories), ordinal (e.g., hierarchical order of categories), interval (distance between members of an order), ratio (interval compared to a meaningful "0"), or cardinal number measurements that count the number of things in a set. Measurements of variables on a nominal scale indicate names or categories, e.g., categories into which sequencing reads fall. Measurements of variables on an ordinal scale result in rankings such as "1st," "2nd," "3rd," etc. Measurements on a ratio scale include, for example, any measurements on a given scale, absolute numbers of reads, normalized or estimated numbers, and statistical measurements such as frequencies, means, medians, standard deviations, or quantiles. Measurements that involve quantification are usually determined at the ratio scale level.

[0034] Culture output can be measured by any method known in the art. For example, growth rate can be measured as a function of culture turbidity. Proteins can be measured, for example, by immunoassays. Enzymes can be measured by enzyme assays. Small molecules can be measured by gas chromatography, HPLC, or mass spectrometry.

[0035] II. Process Referring to Figure 1, optimization of culture conditions for culture output generally involves three stages: In a first stage 100, the genetic stability of the cells in culture is established. In a second stage 200, the culture conditions are optimized for cell growth 210 and / or culture output or a surrogate thereof 220. In a third stage 300, the culture conditions are scaled up and validated. Each of these three stages can be performed independently or together.

[0036] These processes are detailed in Figure 2. The preparation phase can include initial validation and verification runs, where baseline titer and productivity are established. In an initial turbidostat run, baseline genetic stability is established. In the optimization phase, optimized growth parameters are determined in a turbidostat and optimized production parameters are determined using a chemostat. In the validation phase, the titer and productivity improvements of the selected optimized culture conditions are confirmed at scale-up compared to the baseline in fed-batch or batch operations.

[0037] A. Establishing genetic stability The first step in optimization can include determining the genetic stability of the cells in culture. Genetic stability refers to the number of generations that the cells in culture can grow before production of a particular culture output is significantly impaired. It can also be expressed as the duration of culture before production of a particular culture output is significantly impaired. Production of a culture output is said to be significantly impaired if production of the culture output is impaired by at least 20%, at least 25%, or at least 50%.

[0038] To determine genetic stability, cells are established in continuous culture. Once established, the production of culture output is measured serially over time. At some point, the production of culture output will deteriorate. The number of generations, or the time until the production of culture output deteriorates (which can also be measured as generation cycles), represents the stability of the cell line.

[0039] B. Optimization of Culture Conditions Provided herein is a method for optimizing culture conditions in a culture process to generate a culture output. The method includes: determining a set of previous culture outputs for each set of previous culture conditions; testing new culture conditions that are expected to improve the previous culture outputs to determine new culture outputs; and repeating the previous operations until a new culture condition that improves the best previous measurement of the culture output is identified. The step of identifying a new culture condition that generates an improved culture output over the previous culture condition is repeated to result in a culture condition that shows a continuous improvement of the culture output, and can be carried out iteratively to continuously improve the culture condition. The method can continue, for example, until a target measurement of the culture output is achieved, until further improvement of the measurement of the culture output cannot be achieved after several iterations, or until the culture parameter space is exhausted or considered to be fully explored.

[0040] Figure 3 shows the optimization process in detail. A seed culture is established that is used to inoculate the cultures, and continuous culture conditions are used that include an initial set of culture parameters. In the optimization phase, the culture is driven towards a steady state. One or more culture outputs are measured. An operator or agent (e.g., an automated agent using reinforcement learning) interprets the results and decides which parameters to change that are predicted to result in improved performance. The system is perturbed with these parameters and brought back to a steady state. Culture outputs are measured again. This process is repeated until the optimization is complete. At this point, the operator can move to the validation phase.

[0041] 1. Optimization Optimization is a process in which the measured output of a process is improved through multiple iterations (in which experimental parameters are changed). The multiple iterations can be at least 2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 25, 30, 40, 50, 60, 70, 80, 90, or 100. A process is said to be optimized when repeated changes of experimental parameters within the parameter space do not result in improved measured output. Typically, such optimization is achieved after 4-15 iterations when the parameter space contains 2-8 of the parameters being tested. The goal is to identify a minimum or maximum in the multidimensional (culture condition) space. The presence of a local maximum or minimum may imply a process optimized within a limited parameter space. However, a more extensive exploration of the parameter space may lead to the discovery of a more optimal outcome. Therefore, in optimization it is possible to identify local or global maxima or minima over the parameter space.

[0042] In an optimization process, one or more parameters, or variables, may be tested. The parameters to be optimized may be one or more, two or more, three or more, four or more, five or more, six or more, seven or more, or eight or more. The bounds of any parameter are the maximum and minimum values ​​that the parameter can reasonably take. The parameter space is the space bounded by the bounds of all the parameters being tested. Thus, for example, the parameter space of pH and temperature may be the space bounded by pH=2 and pH=10.0; and temperature=15° C. and temperature=50° C. Any pair of numbers within these two ranges is said to be in the parameter space. Exhausting the parameter space would include testing combinations throughout the space. This would include, for example, the space around the extremes of each parameter, as well as combinations between the extremes. Thus, for example, if the pH extremes are pH2 and pH10, and the temperature extremes to be tested are 15° C. and 50° C., an exhaustive search would include measurements at least at about pH2 and about 15° C.; about pH2 and about 50° C.; pH10 and about 15° C.; and pH10 and about 50° C. Within these extremes, multiple combinations around the mid-range (e.g., about pH6 and about 32° C.) would also be tested.

[0043] Optimization of one or more culture outputs is possible. For example, conditions can be optimized for growth rate or for production of culture product. Alternatively, two or more culture outputs can be optimized simultaneously. Culture conditions can be optimized for two or more culture outputs by weighting the culture outputs in the decision process. For example, CO2 production is an indicator of how much nutrients are being converted to product. Thus, a lower CO2 production rate is better. Thus, both CO2 production and product titer can be measured. An optimum may combine both scores, weighting product titer higher than CO2 production rate.

[0044] 2. Iterative testing Using the methods described herein, the culture parameter space can be rapidly and efficiently explored to identify optimal culture conditions for one or more selected culture outputs.

[0045] Up front, the user selects the line of cells to be cultured and the culture output(s) whose measurements will be optimized. The cells are placed in a bioreactor containing culture medium and a steady state is typically established in continuous culture. The desired culture output can be cell growth rate, production of a molecular entity, production cost, or a proxy for these.

[0046] Figure 5 shows that cells are maintained at a steady state of growth over time. The cell concentration is indicated in "AU". If one is trying to optimize cell growth, i.e., the growth rate is the culture output to be optimized, then a turbidostat is used, which can keep the turbidity of the culture, which reflects the cell density, constant. The culture parameters are initially set to bring it to a steady state. This may require, for example, 3-5 or more generations of cells. Some of the culture parameters are the ones whose space is to be tested to identify the optimal growth condition(s). The turbidostat maintains the cell density by removing the culture medium containing the cells from the bioreactor and replacing that volume with fresh medium. Thus, the growth rate is a function of the amount of culture medium removed over time when the culture is at a steady state.

[0047] The speed of repeated testing in continuous culture is shown in Figure 6. Up to 12 different conditions are shown to be tested in one run of continuous culture. In contrast, a single culture condition can be tested in one run of batch culture.

[0048] The process of optimizing culture conditions is iterative. It involves starting with an initial measurement of culture output. The rate of cell growth can be expressed, for example, as the doubling time of cells in culture. The rate of cell growth can be initially tested under multiple different culture conditions. The changes in culture conditions may involve changing only one variable at a time, such as changing temperature or pH, or may involve changing multiple variables, such as changing both temperature and pH. In any case, the initial measurements of culture output under one or more different conditions serve as "priors" against which future culture conditions may be selected.

[0049] A human operator or intelligent agent now selects new culture conditions for testing. Typically, the new culture conditions selected are those that the operator or agent predicts will improve the measured culture output, based on one or more previous measurements of the culture output. Thus, for example, if three different culture conditions, including increasing the culture temperature, result in an increase in culture output, the operator or agent might predict that further increasing the temperature will also further increase the culture output. Thus, the operator or agent may select new culture conditions that further increase the temperature. The culture is then perturbed to reflect the new culture conditions, and the culture output is once again measured.

[0050] Once the culture has reached steady state, a new measurement can be made, or the measurement of culture output can be based on testing the culture after a period of time, or multiple measurements over time can be used to identify a trend, which can then be used to predict the final measured value of culture output.

[0051] The process is repeated, including checking previous measurements of the culture output, predicting changes in culture conditions that are expected to improve the culture output based on these previous values, perturbing the system with the predicted culture conditions, and measuring the culture output of the perturbed system. The number of iterations continues until a specific event is achieved. This can continue until genetic stability of the strain is achieved. That is, iterative testing can be performed for the number of generations before the product output begins to deteriorate due to genetic instability of the cells. The iterations continue until the target culture output is achieved. For example, this can be a constant production rate of the molecular product. One indication that the culture conditions have been optimized is that continuing to change the culture conditions does not result in further improvement in the culture output. It should be understood that the presence of local maxima or minima in the culture parameter space can provide a false indication of global optimization. Therefore, various points can be tested throughout the parameter space to avoid getting stuck in local minimum / maximum regions.

[0052] The initial set of conditions for testing the culture product is informed by the culture conditions found to be optimized for the rate of cell growth. For example, the initial set of parameters can be the same or nearly the same as those identified in the cell growth stage. The initial culture parameters are nearly the same as the previous parameters if each culture parameter is nearly the same as the previous parameter, i.e., within plus or minus 5% of the measurement of the previous culture parameter.

[0053] 7 shows an exemplary set of replicates testing two variables. In this case, the parameters to be varied are pH and temperature. The culture output mu (mu) represents the growth rate of the organism. After several trials, the optimum values ​​of the parameters that maximize mu are identified at approximately pH 7 and temperature approximately 37° C.

[0054] It is also understood that this process may include testing new culture conditions that are not necessarily predicted to improve the culture output in order to provide more data to feed into priors. Such testing may occur at any point in the iterative process. Testing various culture conditions may include a systematic process of systematically varying variables of each culture parameter across the parameter space. Such a process will identify the culture condition(s) that result in the optimal value of the culture output. However, such a process may be slower and less efficient than an intelligent process that uses previous results to predict culture conditions that will improve the culture output.

[0055] Thus, the process can be used to identify culture conditions that result in the highest growth rate for a culture of cells. Additionally, the process can be used to identify culture conditions that are optimal for the production of a product, or that minimize the cost of producing the product.

[0056] In certain embodiments, the culture output measured can be the amount of a particular molecular product produced. In other embodiments, the absolute amount of product produced can be less important than the economics of product production. Thus, for example, the culture output can be the relative cost to produce one unit of molecular product. The value of the output can then take into account the costs of all inputs. This will reflect the added value of the process.

[0057] C. Validation of culture conditions on a large scale After optimal conditions are identified through iterative testing methods in continuous culture as described herein, these conditions can be scaled up and validated. Typically, production of a product in an industrial system is performed at a scale of 10,000 liters to 1,000,000 liters using (fed-)batch reactors. The validation step involves establishing a culture of cells in the (fed-)batch reactor under the selected optimal culture conditions and measuring the culture output in comparison to baseline starting conditions. If the culture output in the (fed-)batch reactor is proven to reasonably correlate with the output under optimal conditions determined in the continuous culture, these conditions can be used in a commercial process to produce the product.

[0058] III. System Also provided herein is a system for optimizing culture conditions for the production of a culture output, comprising a bioreactor for growing cells and a computer including an intelligent agent, the computer being programmed to instruct the bioreactor to establish culture conditions and to predict from previous data culture conditions expected to improve the culture output.

[0059] A. Computer Systems The models provided herein can be implemented on a programmable digital computer.

[0060] 9 shows an exemplary computer system. The computer system 9901 includes a central processing unit (CPU, also referred to herein as "processor" and "computer processor") 9905, which may be a single-core or multi-core processor, or multiple processors for parallel processing. The computer system 9901 also includes memory or memory locations 9910 (e.g., random access memory, read-only memory, flash memory), electronic storage 9915 (e.g., hard disk), communication interface 9920 (e.g., network adapter) for communicating with one or more other systems, and peripheral devices 9925, such as cache, other memory, data storage device, bioreactor, and / or electronic display adapter. The computer-readable memory 9910, storage 9915, interface 9920, and peripheral devices 9925 can communicate with the CPU 9905 via a communication bus (solid lines) such as a motherboard. The storage 9915 can be a data storage device (or data repository) for storing data. The computer system 9901 may be operatively coupled to a computer network ("network") 9930 with the aid of a communication interface 9920. The network 9930 may be the Internet, an Internet and / or an extranet, or an intranet and / or an extranet in communication with the Internet. The network 9930 is, in some cases, a telecommunications network and / or a data network. The network 9930 may include one or more computer servers, enabling distributed computing such as cloud computing.

[0061] The CPU 9905 can execute a series of machine-readable instructions, which may be embodied in a program or software (code). The instructions may be stored in a memory location, such as the computer-readable memory 9910. The instructions are transmitted to the CPU 9905, which can then be programmed or otherwise configured to perform the methods of the present disclosure.

[0062] The storage device 9915 may store files such as drivers, libraries, saved programs, etc. The storage device 9915 may store user data, such as user preferences, log files, video or other images, and user programs. The computer system 9901 may optionally include one or more additional data storage devices that are external to the computer system 9901, such as located on a remote server in communication with the computer system 9901 via an intranet or the Internet.

[0063] The computer system 9901 can communicate with one or more remote computer systems via a network 9930 .

[0064] The methods described herein may be performed by machine (e.g., computer processor) executable code stored in an electronic storage location of the computer system 9901, such as, for example, computer readable memory 9910 or electronic storage device 9915. The machine executable or machine readable code is provided in the form of software. In use, the code may be executed by the processor 9905. In some cases, the code is retrieved from storage device 9915 and stored in memory 9910 for ready access by the processor 9905. In some situations, the electronic storage device 9915 is eliminated and the machine executable instructions are stored in memory 9910. The code may be used to communicate with and instruct electronic devices on the equipment, such as circuit boards 9940, modules, or subsystems.

[0065] The computer system 9901 can communicate with one or more remote computer systems via a network 9930 .

[0066] The machine executable code may be stored in electronic storage devices, such as memory (e.g., read-only memory, random access memory, flash memory) or hard disks. A "storage" type medium may include any or all of the tangible memory of a computer, processor, or the like, or their associated modules (e.g., various semiconductor memories, tape drives, disk drives, etc.), which may provide non-transitory storage for the software programming at any time. All or a portion of the software may at times be communicated via the Internet or various other telecommunications networks.

[0067] The computer system 9901 can include or communicate with an electronic display 9935 that includes a user interface (UI) 9940, e.g., for providing input parameters for the methods described herein. Examples of UIs include, but are not limited to, graphical user interfaces (GUIs) and web-based user interfaces.

[0068] Further provided herein is a computer system that includes a tangible software product that includes code that, when executed, directs the system provided herein to perform the optimization methods described herein.

[0069] Machine learning optimization Machine learning optimization is the process of iteratively improving the accuracy of a machine learning model and decreasing the degree of error. Optimization is measured through a loss function or cost function; this is usually a way of defining the difference between the predicted and actual values ​​of the data. Machine learning models aim to minimize this loss function. In our case, the cost function can be the inverse of the level of production of the culture output. Thus, minimizing the cost function is equivalent to maximizing the production of the culture output. Machine learning optimization methods include, for example, random search, gradient descent, and genetic algorithms.

[0070] A random search, or exhaustive search, is a method that tests points throughout the entire parameter space and identifies a minimum of the cost function based on those results.

[0071] In gradient descent, an arbitrary value for a parameter or set of parameters is selected and tested. Then, neighboring values ​​of the parameter in one direction (e.g., higher or lower) are tested to determine whether they are closer to or further from the minimum. If they are, the next parameter value is selected in the other direction and tested. This process can continue until a local or global minimum is found.

[0072] A genetic or evolutionary algorithm tests multiple sets of parameter values ​​and selects the parameter values ​​that minimize a cost function. Thus, for example, various combinations of pH, temperature, and dissolved oxygen are tested for culture output. Values ​​for each parameter associated with improved culture output are selected, recombined, and tested again to identify better combinations of parameter values.

[0073] One of the techniques of deep learning is reinforcement learning. Reinforcement learning is an aspect of machine learning where an agent learns to act in an environment by performing specific actions and observing the rewards / consequences that result from those actions. A schematic diagram of reinforcement learning is shown in Figure 3. The agent learns to act (a t ) is taken. This is because the state of the environment (S t ) and a reward (R t ) The agent works on the reward maximization hypothesis.

[0074] One version of reinforcement learning is called direct search, which refers to sequential consideration of trial solutions, comparing each one with the "best" obtained so far, along with a strategy for deciding what the next trial solution will be (as a function of the previous results).

[0075] Q-learning is a model-free reinforcement learning algorithm for learning the value of actions in specific states.

[0076] In some variations, the analysis may include implementation of machine learning techniques, including linear and non-linear models, including processes such as classification and regression trees (CART), artificial neural networks such as backpropagation networks, discriminant analysis (e.g., Bayesian classifiers or Fisher analysis), logistic classifiers, and support vector classifiers (e.g., support vector machines).

[0077] Machine learning algorithms for identifying optimal culture conditions can leverage deep learning techniques, which use multiple layers in the learning process.

[0078] Artificial neural networks use a collection of interconnected nodes. A neural network is composed of a node layer, which includes an input layer, one or more hidden layers, and an output layer. Each node, or artificial neuron, connects to other nodes and has an associated weight and threshold. When the output of an individual node exceeds a specified threshold, the node is activated and sends data to the next layer of the network (which consists of a node layer, which includes an input layer, one or more hidden layers, and an output layer). Each node, or artificial neuron, connects to other nodes and has an associated weight and threshold. When the output of an individual node exceeds a specified threshold, the node is activated and sends data to the next layer of the network.

[0079] Exemplary Aspects 1. a) providing a continuous culture of cells in a bioreactor, the cells being cultured under culture conditions comprising a parameter set of culture parameters, the culture of cells producing at least one culture output in a first measurement; b) maintaining the cells in continuous culture; (i) selecting new culture conditions comprising a new set of parameters, the new culture conditions being predicted to improve the measured value of a culture output and being selected based on one or more previous measurements of the culture output; (ii) perturbing the cells with the new culture conditions; and (iii) obtaining new measurements of culture output; and iteratively testing different culture conditions by, wherein the testing is repeated to produce multiple improvements in the measured culture output. 16. A method for iteratively improving culture conditions for at least one culture output in a culture process, comprising: 2. (1) the target output measurement is achieved; or (2) multiple iterations no longer identify culture conditions that improve on previous optimal output measurements; The method of embodiment 1, comprising iteratively testing different culture conditions until 3. The method of embodiment 1, further comprising testing new culture conditions until the exploration of the user-defined parameter space is exhausted. 4. The method of embodiment 1, wherein the selecting is performed using an automated agent. 5. The method of aspect 4, wherein the automated agent uses artificial intelligence methods. 6. The method of embodiment 5, wherein the artificial intelligence method is selected from reinforcement learning, deep learning, Q-learning, and neural nets. 7. The method of embodiment 4, wherein the automated agent uses a direct search method (e.g., Nelder-Mead algorithm, dynamic programming, downhill simplex, simplex algorithm, linear regression, binary search tree, or random walk). 8. The method of embodiment 1, further comprising, prior to operation (b), determining a baseline genetic stability of the cells in culture. 9. The method of embodiment 1, comprising repeating operation (b) at least any of 2, 3, 4, 5, 10, 15, 20, 25, 30, 50, or 100 times (e.g., 4-15 times). 10. The method of embodiment 1, wherein the culture output comprises a weighted score comprising a plurality of different culture outputs. 11.c) selecting, from among the parameter sets, a parameter set that optimizes the measured culture output; growing the cells in a batch culture (e.g., a fed-batch culture) using the selected parameter set; and measuring the culture output. The method of embodiment 1, further comprising: 12. The method of embodiment 1, wherein the volume of the culture is from about 50 mL to about 15 L, for example, from about 500 mL to about 10 L. 13. The method of embodiment 1, wherein the culture output is a specific growth rate. 14. The method of embodiment 1, wherein the culture output is selected from product titer, product yield, specific production rate, volumetric production rate. 15. The method of embodiment 1, wherein the culture output is selected from biomass yield, specific CO2 production rate, specific O2 consumption rate, organic acid profile, metabolite profile, by-product profile, and production economics. 16. The method of embodiment 1, wherein the culture output is selected from polypeptides (e.g., proteins, enzymes, antibodies), organic molecules that are products of synthetic pathways within the cells (e.g., industrial chemicals such as flavors (e.g., vanillin), fragrances (e.g., aldehydes, coumarins, indoles), amino acids, organic acids (e.g., citric acid, lactic acid, acetic acid), alcohols (e.g., ketones such as ethanol, isopropanol, acetone), and fatty acids (e.g., palmitic acid, oleic acid)). 17. The method of embodiment 13, wherein the specific rate is defined as the rate of change of culture output per bacterium per hour. 18. The method of embodiment 1, wherein the cells comprise archaea, prokaryotes and / or eukaryotes. 19. The method of embodiment 1, wherein the cell comprises a fungal cell (e.g., a yeast (e.g., Saccharomyces cerevisiae, Pichia spp., Kuyveromyces spp.) or Aspergillus spp., Rhodoporidium spp., Lipolytica spp., Aspergillus spp., Neurospora spp., Trichoderma spp., Candida spp., Penicillium spp.). 20. The method of embodiment 1, wherein the cells comprise bacterial cells (e.g., Escherichia coli, Bacillus spp, Costridia spp, Streptomyces spp, Pseudomonas spp, Ralstonia spp, Shewanella spp). 21. The method of embodiment 1, wherein the cell comprises an insect cell, an animal cell or a plant cell. 22. The method of embodiment 1, wherein the cell comprises an animal cell (e.g., an arthropod (e.g., insect, shrimp, lobster, crayfish, and crab) and a chordate (e.g., fish, amphibian, reptile, bird (e.g., chicken or turkey), and a mammal (e.g., human or non-human, e.g., cow, lamb, goat, pig, horse, dog, cat, primate)). 23. The method of embodiment 1, wherein the cells include cell lines (e.g., CHO (Chinese Hamster Ovary cells), BHK21 (Baby Hamster Kidney), NS0, Sp2 / 0 mouse cell lines), insect cells (e.g., SP9, Sf9, sf21, S2), tobacco BY-2 cells, rice (Oryza Sativa), and algae cells. 24. The method of embodiment 1, wherein said measurement of culture output is based on cells grown to stationary state. 25. The method of embodiment 1, wherein said measurement of culture output is based on a trend line of growing cells. 26. The method of embodiment 1, wherein said measurements are made between 1 hour and 24 hours after application of the perturbation. 27. The method of embodiment 1, wherein the culturing is carried out in a turbidostat or chemostat. 28. The method of embodiment 1, wherein one or more of the culture parameters include one or more of temperature, pH, dissolved oxygen, carbon feed rate, and nitrogen feed rate. 29. The method of embodiment 1, wherein the one or more culture parameters include nutrient concentrations. 30. One or more culture parameters are not statistically significant for metals (e.g., iron, zinc, cobalt, copper, nickel, manganese, molybdate, selenite and other transition metals), vitamins (e.g., niacin, pyridoxine, riboflavin, pantothenate, aminobenzoic acid(s), thiamine, biotin, cyanocobalamin, folic acid), inducers, salts, nitrogen sparing rate, aeration rate, oxygen sparge rate, carbon dioxide sparing rate, phosphate, sulfate, chloride, acetate, citrate and other anion salts, magnesium, calcium, sodium, potassium, aluminium, potassium, potassium chloride ... The method of embodiment 1, comprising concentrations of ammonium and other cation salts, boric acid, choline, ascorbic acid, lipoic acid, nicotinic acid, inositol and other vitamins, antifoaming agents (e.g., Antiform 204, Antiform A, Antiform C), amino acids (e.g., glutamic acid, leucine, and tryptophan), nucleobases (e.g., adenine, cytosine, thymine, uracil, and guanine), complex nutrients (e.g., yeast extract, peptone, tryptone, and casamino acids), macronutrients, and / or micronutrients. 31. The method of embodiment 1, wherein the one or more culture parameters include a concentration of a cell growth factor, a concentration of CO2, a culture agitation rate, and a concentration of an antibiotic. 32. The method of embodiment 1, wherein the culture parameters include a concentration of a carbon source. 33. The method of embodiment 32, wherein the carbon source is selected from sugar (e.g., glucose, xylose, sucrose, glycerol, or acetate), molasses, malt extract, starch, dextrin, pulp, CO, or CO2. 34. The method of embodiment 1, wherein the culture parameters include a concentration of a nitrogen source. 35. The method of embodiment 34, wherein the nitrogen source is selected from an amino acid or polypeptide, urea, an ammonium salt (e.g., ammonium sulfate, ammonium phosphate, or ammonia), corn steep liquor, yeast extract, peptone, and soybean meal. 36. The method of embodiment 1, wherein the parameter set includes at least any of 1, 2, 3, 4, 5, 6, 7, or 8 (e.g., 2 to 6) culture parameters. 37. The method of embodiment 1, comprising performing 2 to 20 (e.g., 8 to 15) repetitions within a 24 hour period. 38. The method of embodiment 1, comprising inducing activity of a biochemical pathway that produces a target product. 39. The method of embodiment 1, wherein the cells comprise a constitutive or inducible gene expression system. 40. The method of embodiment 39, wherein the gene encodes a commercial product. 41. The method of embodiment 39, wherein the gene encodes an enzyme in a biochemical production pathway of a commercial product. 42. The method of embodiment 1, wherein the product is selected from a recombinant protein, a natural protein (e.g., an enzyme), a nutritional supplement, a cannabinoid, a metabolite, an organic acid(s), a lipid(s), a spore, or a cellular biomass. 43. Before step (b), To determine baseline fermentation process microbial growth, production titers, and productivity in batch or fed-batch reactors The method of embodiment 1, further comprising: 44. The method of embodiment 1, comprising a first and a second optimization process, the first optimization process identifying an optimized parameter set for cell growth and the second optimization process identifying an optimized parameter set for a molecular product. 45. The method of embodiment 44, wherein an optimized parameter set for cell growth informs initial culture conditions for optimizing the molecular product. 46. ​​The method of aspect 1, wherein the initial parameter set comprises a randomized, repeated, or selected parameter set. 47. A system including: a) A vessel configured for continuous culture of cells (e.g., a turbidostat or chemostat); b) one or more sensors that (i) measure one or more culture outputs of cells cultured in the vessel and (ii) transmit the measurements to a computer memory; c) one or more supply means for supplying the formulation (e.g. nutrients) to said container; d)(i) a processor; (ii) a module coupled to the processor, the module comprising: (1) a dataset including, for each of a plurality of culture conditions, measurements of culture parameters and measurements of culture outputs received from sensors and associated with the culture conditions; (2) an automated agent that includes an algorithm for selecting new culture conditions predicted to result in improved measured culture output and based on the data set; and (3) computer-executable instructions for implementing new culture conditions in the system, measuring the culture output, and adding the culture conditions and culture outputs to the data set; A memory including A computer that includes 48. The system of aspect 47, wherein the automated agent uses artificial intelligence methods. 49. The system of embodiment 48, wherein the artificial intelligence method is selected from reinforcement learning, deep learning, Q-learning, and neural nets. 50. The system of aspect 47, wherein the automated agent uses a direct search method (e.g., Nelder-Mead algorithm, dynamic programming, downhill simplex, simplex algorithm, linear regression, binary search tree, or random walk). 51. A) A motor-operated mixer (e.g., an impeller or air-operated agitator) for mixing liquids in a vessel; B) Temperature regulator and temperature sensor; C) a source of acidic and alkaline reagents in communication with the interior of the vessel, and a pH meter; D) An aeration system in communication with the interior of the vessel, and a dissolved oxygen meter; E) a source of nutrients in communication with the interior of the vessel, and an analyzer for measuring the nutrients; F) an exhaust in communication with the interior of the vessel, and an adjustable valve or pump for regulating fluid flow rate from the vessel; G) Baffles within the vessel; H) a sparger and mass flow controller in communication with the interior of the vessel for introducing one or more gases and mixtures thereof; and I) A user interface for communicating instructions to a computer The system of embodiment 47, comprising one or more of the following: 52. The system of embodiment 51, wherein operation of the mixer, temperature regulator, source of acidic and alkaline reagents, source of oxygen, and source of nutrients are under computer control. 53. a) To determine the genetic stability of cells in culture in continuous culture; b) culture conditions including a parameter set of culture parameters for a culture output; (i) optimizing culture conditions for cell growth; and / or (ii) optimizing culture conditions for product output Optimizing by I) providing initial measurements of cell growth or product output under multiple different parameter sets; II) Repeatedly performing the following: A) selecting a new set of parameters that is predicted to improve cell growth or product output based on measurements of cell growth or product output; B) perturbing cells in culture with the new parameter set; and C) measuring cell growth or product output after perturbing the cells; and III) Selecting an optimized set of parameters said optimizing comprising performing a series of operations including: c) The optimized parameter set is i) growing the cells in continuous or fed-batch mode under an optimized set of parameters; and ii) Measuring culture output To verify by The method includes: 54. The method of aspect 53, wherein the selecting is performed by an automated agent. 55. The method of embodiment 53, wherein genetic stability is measured as a function of the number of generations until a measure of cell growth or product output is significantly deteriorated. 56. The method of embodiment 53, wherein optimization of cell growth is performed in a turbidostat and optimization of culture output is performed in a chemostat. 57. The method of embodiment 53, wherein optimization in continuous culture is not carried out for longer than the period of genetic stability. 58. The method of embodiment 53, wherein the validating is carried out in a volume of about 5 liters to about 500 liters. 59. a) providing a continuous culture of cells in a bioreactor, the culture of cells producing at least one culture output; and b) using a machine learning optimization method to iteratively vary and test different culture conditions along with one or more different culture parameters for the production of at least one culture output while maintaining the cells in continuous culture, wherein multiple iterations vary the culture conditions in a direction predicted by the model to improve the production of at least one culture output, and the iterations are continued to produce multiple improvements. The method includes: 60. The method of embodiment 59, wherein the iteration continues until a local maximum in the generation is reached. 61. The method of embodiment 59, wherein the repetition is performed 2 to 25 times, for example 5 to 20 times. 62. The method of embodiment 59, wherein the repetition is performed at least 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25 times. 63. The method of embodiment 59, wherein the iterations are performed to result in at least any of 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24, or 25 improvements. 64. The method of embodiment 59, wherein the machine learning optimization method uses reinforcement learning, gradient descent, a genetic algorithm, or random search. 65. The method of embodiment 59, wherein the multiple iterations include altering culture conditions along multiple culture parameters. 66. The method of embodiment 65, wherein the plurality of culture parameters includes one or more of pH, temperature, dissolved oxygen, carbon feed rate, and nitrogen feed rate. 67. The method of embodiment 59, further comprising culturing the cells under improved or optimized culture conditions to produce a culture product, and recovering the culture product from the culture. 68. The method of embodiment 67, wherein the culture product is produced at a scale of at least any of 500 liters, 1000 liters, 5000 liters, 10,000 liters, 50,000 liters, 100,000 liters, 500,000 liters, or 1,000,000 liters. EXAMPLES

[0080] A continuous turbidostat culture of Escherichia coli cells is performed using a defined medium to optimize the optimal specific growth rate (mu). The two parameters of interest are growth temperature and pH. The range of growth temperature and pH (25 °C-50 °C, 5.5-8.5) are pre-determined before starting the optimization. The reactor is randomly initialized at temperature = 25 °C, pH = 7.5 and entered a steady state at a constant biomass concentration (turbidostat).

[0081] After three volumetric replacements (i.e., three liters of medium are consumed through a turbidostat with a constant 1 liter volume), machine learning or numerical search methods, such as reinforcement learning or the Nelder-Mead method, are performed to search for the optimal point. After the starting conditions, the reactor conditions were changed to temperature = 27 °C, pH = 6.4, then temperature = 29 °C, pH = 7.0; each change resulted in a higher specific growth rate. The specific growth rate is measured at steady state; steady state is achieved at each of these conditions after three volumetric replacements. Over time, the reactor is stopped at the final optimized conditions of temperature = 33 °C, pH = 7.0; this final optimized growth condition results in the highest growth rate. An exemplary result is shown in Figure 7.

[0082] As used herein, the following meanings apply unless otherwise specified: The word "may" is used in a permissive sense (i.e., may) rather than a mandatory sense (i.e., must). The word "include", "including", "includes", etc. means including, but not limited to. The singular forms "a", "an", and "the" include plural referents. Thus, for example, reference to "an element" includes a combination of two or more elements, notwithstanding the use of other terms or phrases that refer to one or more elements, such as "one or more". The phrase "at least one" includes "one", "one or more", "one or more", and "plurality". The term "or" is non-exclusive, i.e., encompasses both "and" and "or", unless otherwise indicated. The term "any of" between a modifier and a sequence means that the modifier modifies every member of the sequence. Thus, for example, the phrase "at least any of 1, 2 or 3" means "at least 1, at least 2, or at least 3." The term "about" refers to a range of plus or minus 5% from the specified numerical value, within the context of the particular use. Thus, for example, "about 100" means between 95 and 105. The term "consisting essentially of" refers to the inclusion of the recited elements and other elements that do not materially affect the basic and novel characteristics of the claimed combination.

[0083] It should be understood that the specification and drawings are not intended to limit the invention to the particular forms disclosed, but on the contrary, are intended to cover all modifications, equivalents, and alternatives falling within the spirit and scope of the invention as defined by the appended claims. Further modifications and alternative embodiments of various aspects of the invention will be apparent to those skilled in the art from this specification. Thus, the specification and drawings should be interpreted as illustrative only and are intended to teach those skilled in the art the general manner of carrying out the invention. It should be understood that the forms of the invention shown and described herein are to be taken as exemplary embodiments. Elements and materials may be substituted for those illustrated and described herein, parts and processes may be reversed or omitted, and certain features of the invention may be utilized independently, all as would be apparent to one skilled in the art after having the benefit of this description of the invention. Changes may be made to the elements described herein without departing from the spirit and scope of the invention as set forth in the following claims.

[0084] All publications, patents, and patent applications mentioned in this specification are herein incorporated by reference to the same extent as if each individual publication, patent, or patent application was specifically and individually indicated to be incorporated by reference.

Claims

1. a) providing a continuous culture of cells in a bioreactor, the cells being cultured under culture conditions comprising a parameter set of culture parameters, the culture of cells producing at least one culture output in a first measurement; b) maintaining the cells in continuous culture; (i) selecting new culture conditions comprising a new set of parameters, the new culture conditions being predicted to improve the measured value of a culture output and being selected based on one or more previous measured values ​​of the culture output; (ii) perturbing the cells with the new culture conditions; and (iii) Obtaining new measurements of culture output and iteratively testing different culture conditions by, wherein the testing is repeated to produce multiple improvements in the measured culture output.

16. A method for iteratively improving culture conditions for at least one culture output in a culture process, comprising:

2. (1) the target output measurement is achieved; or (2) multiple iterations no longer identify culture conditions that improve on previous optimal output measurements; 2. The method of claim 1, further comprising repeatedly testing different culture conditions until

3. 10. The method of claim 1, further comprising testing new culture conditions until the exploration of the user-defined parameter space is exhausted.

4. The method of claim 1 , wherein the selecting is performed using an automated agent.

5. The method of claim 4 , wherein the automated agent uses artificial intelligence methods.

6. 6. The method of claim 5, wherein the artificial intelligence method is selected from reinforcement learning, deep learning, Q-learning, and neural nets.

7. The method of claim 4 , wherein the automated agent uses a direct search method (e.g., Nelder-Mead algorithm, dynamic programming, downhill simplex, simplex algorithm, linear regression, binary search tree, or random walk).

8. 2. The method of claim 1, further comprising determining a baseline genetic stability of the cells in culture prior to the operation (b).

9. 2. The method of claim 1, comprising repeating operation (b) at least any of 2, 3, 4, 5, 10, 15, 20, 25, 30, 50, or 100 times (e.g., 4 to 15 times).

10. 2. The method of claim 1, wherein the culture output comprises a weighted score comprising a plurality of different culture outputs.

11. c) selecting, from among the parameter sets, a parameter set that optimizes the measured culture output; growing the cells in a batch culture (e.g., a fed-batch culture) using the selected parameter set; and measuring the culture output. The method of claim 1, further comprising:

12. 2. The method of claim 1, wherein the volume of the culture is from about 50 mL to about 15 L, such as from about 500 mL to about 10 L.

13. 2. The method of claim 1, wherein the culture output is a specific growth rate.

14. 2. The method of claim 1, wherein the culture output is selected from product titer, product yield, specific production rate, volumetric production rate.

15. Culture output is expressed as biomass yield, specific CO 2 Formation rate, ratio O 2 2. The method of claim 1, wherein the determining factor is selected from consumption rate, organic acid profile, metabolite profile, by-product profile, and production economics.

16. 2. The method of claim 1, wherein the culture output is selected from polypeptides (e.g., proteins, enzymes, antibodies), organic molecules that are products of synthetic pathways within the cells (e.g., industrial chemicals such as flavors (e.g., vanillin), fragrances (e.g., aldehydes, coumarins, indoles), amino acids, organic acids (e.g., citric acid, lactic acid, acetic acid), alcohols (e.g., ketones such as ethanol, isopropanol, acetone), and fatty acids (e.g., palmitic acid, oleic acid)).

17. 14. The method of claim 13, wherein the specific rate is defined as the rate of change of culture output per bacterium per hour.

18. The method of claim 1 , wherein the cells comprise archaea, prokaryotic cells and / or eukaryotic cells.

19. 2. The method of claim 1, wherein the cell comprises a fungal cell (e.g. a yeast (e.g. Saccharomyces cerevisiae, Pichia spp., Kuyveromyces spp.) or Aspergillus spp., Rhodoporidium spp., Lipolytica spp., Aspergillus spp., Neurospora spp., Trichoderma spp., Candida spp., Penicillium spp.).

20. 2. The method of claim 1, wherein the cells comprise bacterial cells (e.g. Escherichia coli, Bacillus spp, Costridia spp, Streptomyces spp, Pseudomonas spp, Ralstonia spp, Shewanella spp).

21. The method of claim 1 , wherein the cell comprises an insect cell, an animal cell or a plant cell.

22. The method of claim 1, wherein the cells include animal cells, such as arthropods (e.g., insects, shrimp, lobsters, crayfish, and crabs) and chordates (e.g., fish, amphibians, reptiles, birds (e.g., chickens or turkeys), and mammals (e.g., human or non-human, e.g., cows, lambs, goats, pigs, horses, dogs, cats, primates).

23. 2. The method of claim 1, wherein the cells include cell lines (e.g., CHO (Chinese Hamster Ovary cells), BHK21 (Baby Hamster Kidney), NS0, Sp2 / 0 mouse cell lines), insect cells (e.g., SP9, Sf9, sf21, S2), tobacco BY-2 cells, rice (Oryza Sativa) and algae cells.

24. 10. The method of claim 1, wherein the measurements of culture output are based on cells grown to steady state.

25. 10. The method of claim 1, wherein the measurement of culture output is based on a trend line of growing cells.

26. 10. The method of claim 1, wherein the measurements are made between 1 hour and 24 hours after application of the perturbation.

27. The method of claim 1, wherein the culturing is carried out in a turbidostat or a chemostat.

28. 2. The method of claim 1, wherein one or more of the culture parameters comprises one or more of temperature, pH, dissolved oxygen, carbon feed rate, and nitrogen feed rate.

29. The method of claim 1 , wherein the one or more culture parameters include nutrient concentrations.

30. One or more culture parameters may be determined to be metals (e.g., iron, zinc, cobalt, copper, nickel, manganese, molybdate, selenite and other transition metals), vitamins (e.g., niacin, pyridoxine, riboflavin, pantothenate, aminobenzoic acid(s), thiamine, biotin, cyanocobalamin, folic acid), inducers, salts, nitrogen sparing rate, aeration rate, oxygen sparge rate, carbon dioxide sparing rate, phosphate, sulfate, chloride, acetate, citrate and other anion salts, magnesium, calcium, sodium, potassium, ammonium, potassium chloride ...

10. The method of claim 1, further comprising the steps of: 1) adding a nutrient solution containing at least one of the following nutrients: vitamins, cations, and other cationic salts, boric acid, choline, ascorbic acid, lipoic acid, nicotinic acid, inositol and other vitamins, antifoaming agents (e.g., Antifoam 204, Antifoam A, Antifoam C), amino acids (e.g., glutamic acid, leucine, and tryptophan), nucleic acid bases (e.g., adenine, cytosine, thymine, uracil, and guanine), complex nutrients (e.g., yeast extract, peptone, tryptone, and casamino acids), macronutrients, and / or micronutrients.

31. One or more culture parameters can be determined by the concentration of cell growth factors, CO 2 2. The method of claim 1, comprising a concentration of the culture medium, a culture agitation rate, and a concentration of the antibiotic.

32. The method of claim 1, wherein the culture parameters include a concentration of a carbon source.

33. The carbon source can be sugar (e.g., glucose, xylose, sucrose, glycerol, or acetate), molasses, malt extract, starch, dextrin, fruit pulp, CO or CO 2 33. The method of claim 32, wherein said compound is selected from the group consisting of:

34. The method of claim 1 , wherein the culture parameters include a concentration of a nitrogen source.

35. 35. The method of claim 34, wherein the nitrogen source is selected from an amino acid or polypeptide, urea, an ammonium salt (e.g., ammonium sulfate, ammonium phosphate, or ammonia), corn steep liquor, yeast extract, peptone, and soybean meal.

36. The method of claim 1, wherein the parameter set comprises at least any of 1, 2, 3, 4, 5, 6, 7, or 8 (e.g., 2 to 6) culture parameters.

37. 10. The method of claim 1, comprising performing 2 to 20 (e.g., 8 to 15) repetitions within a 24 hour period.

38. 10. The method of claim 1, comprising inducing activity of a biochemical pathway that produces a target product.

39. The method of claim 1, wherein the cells comprise a constitutive or inducible gene expression system.

40. 40. The method of claim 39, wherein the gene encodes a commercial product.

41. 40. The method of claim 39, wherein the gene encodes an enzyme in a biochemical production pathway of a commercial product.

42. 2. The method of claim 1, wherein the product is selected from a recombinant protein, a natural protein (e.g., an enzyme), a nutritional supplement, a cannabinoid, a metabolite, an organic acid(s), a lipid(s), a spore, or a cellular biomass.

43. Before step (b), To determine baseline fermentation process microbial growth, production titers, and productivity in batch or fed-batch reactors The method of claim 1, further comprising:

44. 2. The method of claim 1, comprising a first and a second optimization process, the first optimization process identifying an optimized parameter set for cell growth and the second optimization process identifying an optimized parameter set for a molecular product.

45. 45. The method of claim 44, wherein the optimized parameter set for cell growth informs initial culture conditions for optimizing molecular product.

46. The method of claim 1 , wherein the initial parameter set comprises a randomized, repeated, or selected parameter set.

47. Systems including: a) a vessel configured for continuous culture of cells (e.g., a turbidostat or chemostat); b) one or more sensors that (i) measure one or more culture outputs of cells cultured in the vessel and (ii) transmit the measurements to a computer memory; c) one or more supply means for supplying the formulation (e.g. nutrients) to said container; d) (i) a processor; (ii) a module coupled to the processor, the module comprising: (1) a dataset including, for each of a plurality of culture conditions, measurements of culture parameters and measurements of culture outputs received from sensors and associated with the culture conditions; (2) an automated agent that includes an algorithm for selecting new culture conditions predicted to result in an improvement in the measured culture output and based on the data set; and (3) computer-executable instructions for implementing new culture conditions in the system, measuring the culture output, and adding the culture conditions and culture outputs to the data set. A memory including A computer that includes

48. 48. The system of claim 47, wherein the automated agent uses artificial intelligence methods.

49. 49. The system of claim 48, wherein the artificial intelligence method is selected from reinforcement learning, deep learning, Q-learning, and neural nets.

50. 48. The system of claim 47, wherein the automated agent uses a direct search method (e.g., Nelder-Mead algorithm, dynamic programming, downhill simplex, simplex algorithm, linear regression, binary search trees, or random walks).

51. A) a motor-operated mixer (e.g., an impeller or air-operated agitator) for mixing the liquid in the vessel; B) Temperature regulator and temperature sensor; C) a source of acid and alkaline reagents in communication with the interior of the vessel, and a pH meter; D) A vent system in communication with the interior of the vessel, and a dissolved oxygen meter; E) a source of nutrients in communication with the interior of the vessel, and an analyzer for measuring the nutrients; F) an exhaust in communication with the interior of the vessel, and an adjustable valve or pump for regulating fluid flow rate from the vessel; G) Baffles within the vessel; H) a sparger and mass flow controller in communication with the interior of the vessel for introducing one or more gases and mixtures thereof; and I) A user interface for communicating instructions to a computer 48. The system of claim 47, comprising one or more of:

52. 52. The system of claim 51, wherein operation of the mixer, temperature regulator, source of acidic and alkaline reagents, source of oxygen, and source of nutrients are under computer control.

53. a) To determine the genetic stability of cells in culture in continuous culture; b) culture conditions including a parameter set of culture parameters for a culture output; (i) optimizing culture conditions for cell growth; and / or (ii) optimizing culture conditions for product output Optimizing by I) providing initial measurements of cell growth or product output under multiple different parameter sets; II) Iteratively performing the following: A) selecting a new set of parameters that is predicted to improve cell growth or product output based on measurements of cell growth or product output; B) perturbing cells in culture with the new parameter set; and C) measuring cell growth or product output after perturbing the cells; and III) Selecting the optimized parameter set said optimizing comprising performing a series of operations including: c) The optimized parameter set is i) growing the cells in continuous or fed-batch mode under an optimized set of parameters; and ii) Measuring culture output To verify by The method includes:

54. 54. The method of claim 53, wherein the selecting is performed by an automated agent.

55. 54. The method of claim 53, wherein genetic stability is measured as a function of the number of generations until a measure of cell growth or product output is significantly compromised.

56. 54. The method of claim 53, wherein the optimization of cell growth is performed in a turbidostat and the optimization of culture output is performed in a chemostat.

57. 54. The method of claim 53, wherein optimization in continuous culture is performed for no longer than the period of genetic stability.

58. 54. The method of claim 53, wherein the verifying is performed in a volume of about 5 liters to about 500 liters.

59. a) providing a continuous culture of cells in a bioreactor, the culture of cells producing at least one culture output; and b) using a machine learning optimization method to iteratively vary and test different culture conditions along with one or more different culture parameters for the production of at least one culture output while maintaining the cells in continuous culture, wherein multiple iterations vary the culture conditions in a direction predicted by the model to improve the production of at least one culture output, and the iterations are continued to produce multiple improvements. The method includes:

60. 60. The method of claim 59, wherein the iteration continues until a local maximum of the generation is reached.

61. 60. The method of claim 59, wherein the iterations are performed 2 to 25 times, such as 5 to 20 times.

62. 60. The method of claim 59, wherein the repetition is performed at least 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or 25 times.

63. 60. The method of claim 59, wherein the iterations are performed to result in any of at least 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21, 22, 23, 24 or 25 improvements.

64. 60. The method of claim 59, wherein the machine learning optimization method uses reinforcement learning, gradient descent, a genetic algorithm, or random search.

65. 60. The method of claim 59, wherein the multiple iterations comprise varying culture conditions along multiple culture parameters.

66. 66. The method of claim 65, wherein the plurality of culture parameters comprises one or more of pH, temperature, dissolved oxygen, carbon feed rate, and nitrogen feed rate.

67. 60. The method of claim 59, further comprising culturing the cells under improved or optimized culture conditions to produce a culture product, and recovering the culture product from the culture.

68. 68. The method of claim 67, wherein the culture product is produced at a scale of at least any of 500 liters, 1000 liters, 5000 liters, 10,000 liters, 50,000 liters, 100,000 liters, 500,000 liters or 1,000,000 liters.

Images