2. Calculator development
For bioprocesses, a plant can be divided into upstream and downstream areas as well as process utilities. Upstream areas are related to fermentation, while downstream areas are related to product purification. Process utilities are required and can be estimated for both upstream and downstream areas. Any TEA usually begins with, and is specific to, the details of the process being modeled, which historically would importantly include the target molecule being produced, the bioprocess performance, and the details of the downstream purification. This complexity can make a generalized TEA tool difficult to implement. As a result, I first sought to develop a generalized, yet meaningful, bioprocess model to enable reasonable cost estimates that are useful early in the strain and process development cycle. The approach taken is illustrated in Fig. 1. Briefly, the Bioprocess TEA Calculator first accounts for the conversion of glucose to the product of interest (Fig. 1a), calculating and accounting for theoretical conversion stoichiometry. Subsequently (Fig. 1b), an aerobic fermentation is modeled, wherein biomass growth is coupled to biosynthesis of the target chemical, given user defined performance metrics. This fermentation model is used for a detailed estimate of the operating (OPEX) and capital (CAPEX) costs of the bioprocess (Fig. 1c), subject to user defined targets for plant performance (plant capacity for example) and downstream purification. As downstream purification (DSP) methods depend on the physical and chemical properties of the product, and other factors, like purity requirements and the fermentation byproduct profile, the current calculator only performs detailed estimates for primary cell removal, with remaining DSP costs estimated as a percentage of overall process costs. Effectively, the bioprocess facility can be thought of as being modeled as two facilities or processes. The first, a more detailed model of a fermentation facility producing clarified or cell free fermentation broth, and a second more crudely modeled purification facility. The DSP is treated as a black box producing a purified chemical product from the clarified broth input. For simplicity, the costs of the DSP are modeled as a user defined fraction of the overall facility capital and operating costs. These DSP estimates can be altered by the user, either informed by a conceptual DSP process or by more general heuristics as discussed below.
Lastly, estimates of OPEX and CAPEX are used along with financial inputs to perform a discounted cash flow analysis (Fig. 1d) leading to estimated commercial outcomes, including the minimum selling price needed to achieve a target margin as well the net present value (NPV), and internal rate of return (IRR) for the project, based on the user defined target selling price.
2.1. Chemistry
As illustrated in Fig. 1, the first input into the calculator is the target chemical product. Currently, the calculator accepts molecules with carbon, hydrogen, oxygen and nitrogen atoms. While many algorithms can be used to balance chemical equations (Herndon, 1997), I chose to deploy a metabolically informed approach. Firstly, the problem is viewed from the perspective of the basic chemical building blocks used for biosynthesis, namely CO2 and H2. Glucose can be thought of as a combination of these two building blocks and specifically glucose provides 2 H2 equivalents per carbon (CO2). Similarly, any potential organic chemical can be broken into carbon (CO2) and reducing equivalents (H2). Unless a chemical has the same H2/CO2 balance as glucose (an H2/CO2 ratio of 2) either carbon (CO2) or reducing equivalents (H2) will be the limiting reagent and so there are fundamentally three classes of "CHO" containing chemicals when considering glucose as a substrate. A chemical is either more reduced than glucose (an H2/CO2 ratio >2, for example ethanol or butanol), neutral with respect to glucose (an H2/CO2 ratio = 2, for example lactate), or more oxidized than glucose (an H2/CO2 ratio < 2, for example acetate or malonate). As Equations Equation 1, Equation 2, Equation 3, Equation 4, Equation 5, Equation 6) illustrate, each of these cases leads to a different chemical equation that needs to be balanced. Algorithmically, the conversion of a given chemical input to H2 and CO2 is first considered, and the H2/CO2 ratio calculated, after which the appropriate final conversion equation is balanced. All of these equations can be balanced using simple algebra as outlined in Supplemental Materials Section 1. In the case of chemical products containing nitrogen a similar approach is used, except that all nitrogen atoms are assumed to originate from ammonia (NH3) which also brings reducing equivalents. This is consistent with basic biochemistry.
equation Equation 1
equation Equation 2
equation Equation 3
equation Equation 4
equation Equation 5
equation Equation 6
These calculations enable an estimate of the theoretical yield for the production of a given chemical from glucose. This calculation is based on the potential theoretical chemical conversion and does not assume a particular biosynthetic pathway or account for thermodynamics. Using a particular biochemical pathway may result in a theoretical pathway yield lower than the theoretical chemical yield. Also, while certain conversions may be stoichiometrically possible they may not be energetically feasible. Additionally, yield coefficients for ammonia and oxygen, used in calculations in the fermentation model below, are calculated at this step. These yield coefficients are essentially the grams of product per gram of ammonia (or oxygen). Due to the details of the stoichiometric calculations, the current version of the calculator assumes glucose (dextrose) as the carbon and electron source for the production of any given target chemical. Future improvements may include additional feedstocks.
2.2. Fermentation model
With the theoretical chemistry defined, I developed a basic model for aerobic fermentation. This model takes into account user defined inputs such as the target titer, volumetric production rate and percentage of theoretical yield as illustrated in Fig. 1. For more details regarding the fermentation model, please refer to Supplemental Materials Section 2. In this model the percentage of theoretical yield is based on the theoretical chemical yield described above. The user should account for differences in yield attributable to a lower theoretical pathway yield for any specific biochemical pathway. For simplicity, the calculator assumes growth associated production and logistic growth. The target yield along with the target titer define the final amount of product and dictate the available sugar (not used for product synthesis) which can be used for biomass growth, as described in Fig. 2. Biomass is modeled as a byproduct chemical (C3.85H6.69O1.78N) with a molecular weight of 95.37 g/mol, which is produced from glucose at 80% of theoretical yield. (Battley, 1987; Grosz and Stephanopoulos, 1983). As a result the user input titer and percent of theoretical yield are used to estimate the final biomass levels at the end of fermentation.
The total fermentation time is calculated from the titer and average volumetric production rate. The starting biomass level is assumed to be 1% of the final biomass (based on a 1% inoculum of a similarly dense seed culture). As mentioned a logistic growth rate is assumed and calculated based on final and initial biomass levels and the fermentation time. Finally the product to cell ratio, a measure of growth associated product biosynthesis, is calculated. This ratio is in effect the gram of product made per gram of biomass as the cells grow. This can be used to estimate a logistic product formation rate.
Together the above calculations lead to a fermentation time course, which can be used for basic material and energy balances and to estimate process requirements including raw materials, oxygen demand and oxygen transfer rates and cooling requirements, which in turn can be used with other inputs to estimate operating and capital costs.
2.3. Downstream purification
Currently, the calculator specifically estimates the costs associated with primary cell removal, which is modeled as centrifugation. These costs are standard in most bioprocesses, whether the product is in the supernatant or the cell pellet. However, the calculator currently only enables the user to estimate additional downstream purification operating and capital costs as a fraction of total costs without any detailed estimates of these unit operations. Importantly, the current estimates assume a single byproduct: carbon dioxide. All dextrose not going to produce biomass or product is completely oxidized to carbon dioxide. In most cases this likely overestimates the oxygen requirements of the process and represents a worse case scenario with respect to aeration, however this also represents a best case scenario with respect to DSP. Real world downstream purification processes must often deal with the removal of several unwanted organic byproducts, which are a function of the product and biosynthetic pathway, as well as strain and process conditions deployed.
Capital costs associated with DSP can range widely depending on the complexity of the DSP and number of unit operations required. Simple distillation, such as is used for ethanol, may be on the lower end (~20%), whereas DSPs requiring chromatographic separations or electrodialysis would be anticipated to account for 50–70% of total capital costs (Gunukula and Anex, 2017; Bechthold et al., 2008; Cheng et al., 2012; Xiu and Zeng, 2008). Operating costs for DSP are often a smaller fraction of the total percentage of OPEX, and I recommend from 20% at the low end to 50% at the high end (Straathof, 2011). Fig. 3, adapted from data presented by Straathof (2011) relates the reported percent of production costs attributed to downstream purification for a variety of molecules. These data can be used for rough estimates of DSP costs for different types of products. Additionally it is worth noting that if the fermentation process producing clarified broth is not financially attractive with a low cost DSP, the DSP details are likely not a critical priority until the fermentation process itself has been further optimized. When DSP details are critical for a project to advance, additional models, beyond this calculator, would need to be built.
2.4. Operating and capital cost estimates
The current version of the Bioprocess TEA Calculator assumes the construction of a new facility in a Build-Own-Operate (BOO) Financial model. Estimated operating and capital costs assume a general plant configuration as illustrated in Fig. 4. Briefly the facility can be broken down into the following 6 key areas including 1) Main fermentation 2) Seed fermentation, 3) Primary cell removal, 4) Process utilities, 5) Downstream purification, and 6) associated infrastructure (warehousing and office buildings, site development etc). In this brownfield design utilities for the process are modeled but main gas lines and electrical supply are provided to the facility. Additionally, while waste is initially heated and pH adjusted, a wastewater treatment plant is not considered in this design, and would need to be provided via colocation. The heating of the waste biomass (at 60 °C) is to kill any surviving bacterial cells prior to transfer to a wastewater treatment facility. This is a conservative step that may not be needed for many microbial processes.
Operating costs are estimated prior to capital costs, as many of the capital estimates depend on the OPEX. Operating costs estimates include variable costs such as raw materials and utilities as well as fixed operating costs including labor and additional fixed costs. For more details on specific calculations underlying estimates refer to Supplemental Materials Sections 3 Use cases and limitations, 4 A technoeconomic analysis of an aerobic bioprocess for diethyl malonate production. The calculator estimates the raw materials and process utilities needed given the target plant capacity and fermentation performance as well as other user defined inputs. Raw materials and utilities estimated include: water, glucose, ammonia, and the costs of CIP (clean in place) reagents, electricity, steam and compressed air. Glucose costs are estimated based on the user defined glucose price as well as the overall fermentation yield and yield in the DSP. First the amount of glucose needed for product biosynthesis and biomass production are calculated as described above to give a target fermentation yield. The fermentation yield together with the overall DSP yield (as input by the user) defines the total glucose needed. The default glucose pricing is at $0.18/lb which corresponds to bulk 2020 pricing for DE95 corn syrup, which is usually >650 g/L dextrose. Lower "over the fence" prices may be achievable when co-locating a facility with a sugar mill. Ammonia costs are estimated similarly to glucose costs.