Optimizing feature selection with gradient boosting machines in PLS regression for predicting moisture and protein in multi-country corn kernels via NIR spectroscopy.

Differences in moisture and protein content impact both nutritional value and processing efficiency of corn kernels. Near-infrared (NIR) spectroscopy can be used to estimate kernel composition, but models trained on a few environments may underestimate error rates and bias. We assembled corn samples from diverse international environments and used NIR with chemometrics and partial least squares regression (PLSR) to determine moisture and protein. The potential of five feature selection methods to improve prediction accuracy was assessed by extracting sensitive wavelengths. Gradient boosting machines (GBMs), particularly CatBoost and LightGBM, were found to effectively select crucial wavelengths for moisture (1409, 1900, 1908, 1932, 1953, 2174 nm) and protein (887, 1212, 1705, 1891, 2097, 2456 nm). SHAP plots highlighted significant wavelength contributions to model prediction. These results illustrate GBMs' effectiveness in feature engineering for agricultural and food sector applications, including developing multi-country global calibration models for moisture and protein in corn kernels.

Global calibration for non-targeted fraud detection in quinoa flour using portable hyperspectral imaging and chemometrics.

Quinoa is one of the highest nutritious grains, and global consumption of quinoa flour has increased as people pay more attention to health. Due to its high value, quinoa flour is susceptible to adulteration. Cross-contamination between quinoa flour and other flour can be easily neglected due to their highly similar appearance. Therefore, detecting adulteration in quinoa flour is important to consumers, industries, and regulatory agencies. In this study, portable hyperspectral imaging in the visible near-infrared (VNIR) spectral range (400-1000 nm) was applied as a rapid tool to detect adulteration in quinoa flour. Quinoa flour was adulterated with wheat, rice, soybean, and corn in the range of 0-98% with 2% increments. Partial least squares regression (PLSR) models were developed, and the best model for detecting the % authentic flour (quinoa) was obtained by the raw spectral data with R2p of 0.99, RMSEP of 3.08%, RPD of 8.77, and RER of 25.32. The model was improved, by selecting only 13 wavelengths using bootstrapping soft shrinkage (BOSS), to R2p of 0.99, RMSEP of 2.93%, RPD of 9.18, and RER of 26.60. A visualization map was also generated to predict the level of quinoa in the adulterated samples. The results of this study demonstrate the ability of VNIR hyperspectral imaging for adulteration detection in quinoa flour as an alternative to the complicated traditional method.

Process design and techno-economic analysis of 2'-fucosyllactose enriched distiller's dried grains with solubles production in dry grind ethanol process using genetically engineered Saccharomyces cerevisiae.

2'-fucosyllactose (2'-FL) has been linked positively with piglet gut health. Genetically engineered Saccharomyces cerevisiae strains producing 2'-FL can be used in the dry grind process to enrich Distiller's dried grains with solubles (DDGS) with 2'-FL and supplement swine diets with 2'-FL. The objectives of our study were to modify dry grind ethanol process for 2'-FL enriched DDGS production and evaluate the techno-economic feasibility of the process. Concentrations of 19.8 g 2'-FL/kg dry DDGS were achieved in the dry grind process using engineered strain without negatively affecting the ethanol yield. Process models for conventional and modified dry grind processes producing 2'-FL enriched DDGS (1150 MT corn/day capacity) were developed using SuperPro Designer. Capital and ethanol production costs for modified dry grind processes were higher than the conventional process. The internal rate of return for the modified processes was higher than the conventional process for $300/MT 2'-FL enriched DDGS selling price.

Improving ethanol yields with deacetylated and two-stage pretreated corn stover and sugarcane bagasse by blending commercial xylose-fermenting and wild type Saccharomyces yeast.

Corn stover and sugarcane bagasse are the most widely available agriculture processing biomass and could serve as feedstocks for production of biofuel. In this study, three different technologies are combined to develop a more efficient conversion process for each of these feedstocks. The three technologies are diluted alkaline deacetylation process, combined thermochemical and mechanical shear pretreatment, and fermentation using a combined inoculum of two commercial Saccharomyces yeast strains. The two yeast strains used were a non-GMO and GMO strain engineered for xylose fermentation. The final ethanol concentrations obtained were 35.7 g/L from deacetylated corn stover and 32.9 g/L from sugarcane bagasse. Blending the two yeast reduced residual xylose content from 1.24 g/L to 0.48 g/L and increased ethanol production by 6.5% compared to solely using the C5/C6 yeast. The optimized yeast blend also lowered the amount of C5/C6 yeast required for inoculation by 80%.

Increasing ethanol yield through fiber conversion in corn dry grind process.

Conversion of corn fiber to ethanol in the dry grind process could increase ethanol yields, reduce downstream processing costs and improve overall process profitability. This work investigates the in-situ conversion of corn fiber into ethanol (cellulase addition during simultaneous saccharification and fermentation) during dry grind process. Addition of 30 FPU/g fiber cellulase resulted in 4.6% increase in ethanol yield compared to the conventional process. Use of excess cellulase (120 FPU/g fiber) resulted in incomplete fermentation and lower ethanol yield compared to the conventional process. Multiple factors including high concentrations of ethanol and phenolic compounds were responsible for yeast stress and incomplete fermentation in excess cellulase experiments.

High-throughput, Microscale Protocol for the Analysis of Processing Parameters and Nutritional Qualities in Maize (Zea mays L.).

Maize is an important grain crop in the United States and worldwide. However, maize grain must be processed prior to human consumption. Furthermore, whole grain composition and processing characteristics vary among maize hybrids and can impact the quality of the final processed product. Therefore, in order to produce healthier processed food products from maize, it is necessary to know how to optimize processing parameters for particular sets of germplasm to account for these differences in grain composition and processing characteristics. This includes a better understanding of how current processing techniques impact the nutritional quality of the final processed food product. Here, we describe a microscale protocol that both simulates the processing pipeline to produce cornflakes from large flaking grits and allows for the processing of multiple grain samples simultaneously. The flaking grits, the intermediate processed products, or final processed product, as well as the corn grain itself, can be analyzed for nutritional content as part of a high-throughput analytical pipeline. This procedure was developed specifically for incorporation into a maize breeding research program, and it can be modified for other grain crops. We provide an example of the analysis of insoluble-bound ferulic acid and p-coumaric acid content in maize. Samples were taken at five different processing stages. We demonstrate that sampling can take place at multiple stages during microscale processing, that the processing technique can be utilized in the context of a specialized maize breeding program, and that, in our example, most of the nutritional content was lost during food product processing.

Fermentation of undetoxified sugarcane bagasse hydrolyzates using a two stage hydrothermal and mechanical refining pretreatment.

In this study, liquid hot water pretreatment was combined with disk milling for pretreatment of sugarcane bagasse. Sugarcane bagasse was pretreated using liquid hot water (LHW) at 140-180 °C for 10 min (20% w/w solids content) and then disk milled. Disk milling improved glucose release 41-177% and ethanol production from glucose/xylose cofermentation by 80% compared to only using LHW pretreatment. The highest ethanol conversion efficiency achieved was 94%, which was observed when bagasse was treated at 180 °C with LHW and disk milled. However, a small amount of residual xylose (3 g/L) was indicative that further improvement could be achieved to increase ethanol production.

Changes in Phenolic Acid Content in Maize during Food Product Processing.

The notion that many nutrients and beneficial phytochemicals in maize are lost due to food product processing is common, but this has not been studied in detail for the phenolic acids. Information regarding changes in phenolic acid content throughout processing is highly valuable because some phenolic acids are chemopreventive agents of aging-related diseases. It is unknown when and why these changes in phenolic acid content might occur during processing, whether some maize genotypes might be more resistant to processing induced changes in phenolic acid content than other genotypes, or if processing affects the bioavailability of phenolic acids in maize-based food products. For this study, a laboratory-scale processing protocol was developed and used to process whole maize kernels into toasted cornflakes. High-throughput microscale wet-lab analyses were applied to determine the concentrations of soluble and insoluble-bound phenolic acids in samples of grain, three intermediate processing stages, and toasted cornflakes obtained from 12 ex-PVP maize inbreds and seven hybrids. In the grain, insoluble-bound ferulic acid was the most common phenolic acid, followed by insoluble-bound p-coumaric acid and soluble cinnamic acid, a precursor to the phenolic acids. Notably, the ferulic acid content was approximately 1950 µg/g, more than ten-times the concentration of many fruits and vegetables. Processing reduced the content of the phenolic acids regardless of the genotype. Most changes occurred during dry milling due to the removal of the bran. The concentration of bioavailable soluble ferulic and p-coumaric acid increased negligibly due to thermal stresses. Therefore, the current dry milling based processing techniques used to manufacture many maize-based foods, including breakfast cereals, are not conducive for increasing the content of bioavailable phenolics in processed maize food products. This suggests that while maize is an excellent source of phenolics, alternative or complementary processing methods must be developed before this nutritional resource can be utilized.

Impact of disk milling on corn stover pretreated at commercial scale.

In cellulosic biofuel production, chemical pretreatment performed at laboratory or pilot scale, followed by mechanical refining, has been demonstrated to be effective to increase feedstock enzyme digestibility. To take the combined pretreatment process one step closer to commercialization, disk milling was performed with commercially pretreated corn stover. Dilute acid pretreated samples with combined severity factors (cSF) of 0.09 (DA09) and 0.43 (DA43) were obtained from a commercial plant. Effects of pretreatment conditions (DA09 and DA43), milling cycles (0, 3, 9, and 15) and enzyme dosages (7.8, 15.6 and 31.2mgcellulase/g dry biomass) were evaluated. Milling improved glucose yields by 0.7 to 1.2-fold. Higher enzyme dosages enhanced sugar yields. Milling was more effective to improve glucose yields, while enzyme dosage was more effective to improve xylose yields. However, dilute acid pretreatment condition was the most important factor to increase final sugar yields compared to milling cycles and enzyme dosages.

Improvement of sugar yields from corn stover using sequential hot water pretreatment and disk milling.

Efficient pretreatment is essential for economic conversion of lignocellulosic feedstocks into monosaccharides for biofuel production. To realize high sugar yields with low inhibitor concentrations, hot water or dilute acid pretreatment followed by disk milling is proposed. Corn stover at 20% solids was pretreated with hot water at 160-200°C for 4-8min with and without subsequent milling. Hot water pretreatment and disk milling acted synergistically to improve glucose and xylose yields by 89% and 134%, respectively, compared to hot water pretreatment alone. Hot water pretreated (180°C for 4min) and milled samples had the highest glucose and xylose yields among all hot water pretreated and milled samples, which were comparable to samples pretreated with 0.55% dilute acid at 160°C for 4min. However, samples pretreated with 1% dilute acid at 150°C for 4min and disk milled had the highest observed glucose (87.3%) and xylose yields (83.4%).

Miscanthus×giganteus xylooligosaccharides: Purification and fermentation.

A procedure was developed to recover xylooligosaccharides (XOS) from Miscanthus×giganteus (M×G) hydrolyzate. M×G hydrolyzate was prepared using autohydrolysis, and XOS rich fractions were acquired using activated carbon adsorption and stepwise ethanol elution. The combined XOS fractions were purified using a series of ion exchange resin treatments. The end product, M×G XOS, had 89.1% (w/w) total substituted oligosaccharides (TSOS) composed of arabinose, glucose, xylose and acetyl group. Bifidobacterium adolescentis and Bifidobacterium catenulatum (health promoting bacteria) were cultured in vitro on M×G XOS and a commercial XOS source, which was used as a comparison. B. adolescentis grew to a higher cell density than B. catenulatum in both XOS cultures. Total xylose consumption for B. adolescentis was 84.1 and 84.8%, respectively for M×G and commercial XOS cultures; and for B. catenulatum was 76.6 and 73.6%, respectively. The xylobiose (X2), xylotriose (X3) and xylotetraose (X4) were almost utilized for both strains. Acetic and lactic acids were the major fermentation products of the XOS cultures.

In Vitro Fermentation of Xylooligosaccharides Produced from Miscanthus × giganteus by Human Fecal Microbiota.

Purified xylooligosaccharides from Miscanthus × giganteus (M×G XOS) were used in an in vitro fermentation experiment inoculated with human fecal microbiota. A commercial XOS product and pectin were used as controls. Decreases in pH by 2.3, 2.4, and 2.0 units and production of short-chain fatty acids (SCFA; acetic acid, 7764.2, 6664.1, and 6387.9 µmol/g; propionic acid, 1006.7, 1089.5, and 661.5 µmol/g; and butyric acid, 955.5, 1252.9, and 917.7 µmol/g) were observed in M×G XOS, commercial XOS, and pectin medium after 12 h of fermentation, respectively. Titers of Bifidobacterium spp., Lactobacillus spp., and Escherichia coli increased when fed all three substrates as monitored by qPCR. There was no significant trend for Clostridium perfringens. During fermentation, M×G XOS was statistically equivalent in performance to the commercial XOS sample as measured by culture acidification and growth of health-promoting bacteria and resulted in the highest SCFA production among the three substrates.

Land usage attributed to corn ethanol production in the United States: sensitivity to technological advances in corn grain yield, ethanol conversion, and co-product utilization.

BACKGROUND: Although the system for producing yellow corn grain is well established in the US, its role among other biofeedstock alternatives to petroleum-based energy sources has to be balanced with its predominant purpose for food and feed as well as economics, land use, and environmental stewardship. We model land usage attributed to corn ethanol production in the US to evaluate the effects of anticipated technological change in corn grain production, ethanol processing, and livestock feeding through a multi-disciplinary approach. Seven scenarios are evaluated: four considering the impact of technological advances on corn grain production, two focused on improved efficiencies in ethanol processing, and one reflecting greater use of ethanol co-products (that is, distillers dried grains with solubles) in diets for dairy cattle, pigs, and poultry. For each scenario, land area attributed to corn ethanol production is estimated for three time horizons: 2011 (current), the time period at which the 15 billion gallon cap for corn ethanol as per the Renewable Fuel Standard is achieved, and 2026 (15 years out). RESULTS: Although 40.5% of corn grain was channeled to ethanol processing in 2011, only 25% of US corn acreage was attributable to ethanol when accounting for feed co-product utilization. By 2026, land area attributed to corn ethanol production is reduced to 11% to 19% depending on the corn grain yield level associated with the four corn production scenarios, considering oil replacement associated with the soybean meal substituted in livestock diets with distillers dried grains with solubles. Efficiencies in ethanol processing, although producing more ethanol per bushel of processed corn, result in less co-products and therefore less offset of corn acreage. Shifting the use of distillers dried grains with solubles in feed to dairy cattle, pigs, and poultry substantially reduces land area attributed to corn ethanol production. However, because distillers dried grains with solubles substitutes at a higher rate for soybean meal, oil replacement requirements intensify and positively feedback to elevate estimates of land usage. CONCLUSIONS: Accounting for anticipated technological changes in the corn ethanol system is important for understanding the associated land base ascribed, and may aid in calibrating parameters for land use models in biofuel life-cycle analyses.

Vacuum stripping of ethanol during high solids fermentation of corn.

In corn-ethanol industry, yeast stress inducing glucose concentrations produced during liquefaction and subsequent high ethanol concentrations produced during fermentation restrict slurry solids to 32 % w/w. These limits were circumvented by combining two novel technologies: (1) granular starch hydrolyzing enzyme (GSHE) to break down starch simultaneously with fermentation and (2) vacuum stripping to remove ethanol. A vacuum stripping system was constructed and applied to fermentations at 30, 40, and 45 % solids. As solids increased from 30 to 40 %, ethanol yield decreased from 0.35 to 0.29 L/kg. Ethanol yield from 45 % solids was only 0.18 L/kg. An improvement was conducted by increasing enzyme dose from 0.25 to 0.75 g/g corn and reducing yeast inoculum by half. After improvement, ethanol yield from 40 % solids vacuum treatment increased to 0.36 L/kg, comparable to ethanol yield from 30 % solids (control).

Autohydrolysis of Miscanthus x giganteus for the production of xylooligosaccharides (XOS): kinetics, characterization and recovery.

The optima conditions of production and purification of xylooligosaccharides (XOS) from Miscanthus x giganteus (MxG) were investigated. Using autohydrolysis, XOS were produced at 160, 180 and 200°C at 60, 20 and 5min, respectively. XOS yield up to 13.5% (w/w) of initial biomass and 69.2% (w/w) of xylan were achieved. Results from HPAEC-PAD analysis revealed that X1-X9 sugar oligomers were produced. Higher temperature and longer reaction time resulted in lower product molecular weight. The three optimum conditions had similar degrees of polymerization XOS. Using 10% activated carbon (w/v) with ethanol/water elution recovered 47.9% (w/w) of XOS from pretreated liquid phase. The XOS could be fractionated by degree of polymerization according to ethanol concentration in the ethanol/water elution. Most of the XOS were washed out in 30% and 50% ethanol/water (v/v) fractions. Recoveries of 91.8% xylobiose, 86.9% xylotriose, 66.3% xylotetrose, 56.2% xylopentose and 48.9% xylohexaose were observed in XOS.

Contamination issues in a continuous ethanol production corn wet milling facility.

Low ethanol yields and poor yeast viability were investigated at a continuous ethanol production corn wet milling facility. Using starch slurries and recycle streams from a commercial ethanol facility, laboratory hydrolysates were prepared by reproducing starch liquefaction and saccharification steps in the laboratory. Fermentations with hydrolysates prepared in the laboratory were compared with plant hydrolysates for final ethanol concentrations and total yeast counts. Fermentation controls were prepared using hydrolysates (plant and laboratory) that were not inoculated with yeast. Hydrolysates prepared in the laboratory resulted in higher final ethanol concentrations (15.8 % v/v) than plant hydrolysate (13.4 % v/v). Uninoculated controls resulted in ethanol production from both laboratory (12.2 % v/v) and plant hydrolysates (13.7 % v/v), indicating the presence of a contaminating microorganism. Yeast colony counts on cycloheximide and virginiamycin plates confirmed the presence of a contaminant. DNA sequencing and fingerprinting studies also indicated a number of dissimilar communities in samples obtained from fermentors, coolers, saccharification tanks, and thin stillage.

A simultaneous saccharification and fermentation model for dynamic growth environments.

Many mathematical models by researchers have been formulated for Saccharomyces cerevisiae which is the common yeast strain used in modern distilleries. A cybernetic model that can account for varying concentrations of glucose, ethanol and organic acids on yeast cell growth dynamics does not exist. A cybernetic model, consisting of 4 reactions and 11 metabolites simulating yeast metabolism, was developed. The effects of variables such as temperature, pH, organic acids, initial inoculum levels and initial glucose concentration were incorporated into the model. Further, substrate and product inhibitions were included. The model simulations over a range of variables agreed with hypothesized trends and to observations from other researchers. Simulations converged to expected results and exhibited continuity in predictions for all ranges of variables simulated. The cybernetic model did not exhibit instability under any conditions simulated.

Design and evaluation of an optimal controller for simultaneous saccharification and fermentation process.

Ethanol from corn is produced using dry grind corn process in which simultaneous saccharification and fermentation (SSF) is one of the most critical unit operations. In this work an optimal controller based on a previously validated SSF model was developed by formulating the SSF process as a Bolza problem and using gradient descent methods. Validation experiments were performed to evaluate the performance of optimal controller under different process disturbances that are likely to occur in practice. Use of optimal control algorithm for the SSF process resulted in lower peak glucose concentration, similar ethanol yields (13.38±0.36% v/v and 13.50±0.15% v/v for optimally controlled and baseline experiments, respectively). Optimal controller improved final ethanol concentrations as compared to process without optimal controller under conditions of temperature (13.35±1.28 and 12.52±1.19% v/v for optimal and no optimal control, respectively) and pH disturbances (12.65±0.74 and 11.86±0.49% v/v for optimal and no optimal control, respectively). Cost savings due to lower enzyme usage and reduced cooling requirement were estimated to be up to $1 million for a 151 million L/yr (40 million gal/yr) dry grind plant.

Starch hydrolysis modeling: application to fuel ethanol production.

Efficiency of the starch hydrolysis in the dry grind corn process is a determining factor for overall conversion of starch to ethanol. A model, based on a molecular approach, was developed to simulate structure and hydrolysis of starch. Starch structure was modeled based on a cluster model of amylopectin. Enzymatic hydrolysis of amylose and amylopectin was modeled using a Monte Carlo simulation method. The model included the effects of process variables such as temperature, pH, enzyme activity and enzyme dose. Pure starches from wet milled waxy and high-amylose corn hybrids and ground yellow dent corn were hydrolyzed to validate the model. Standard deviations in the model predictions for glucose concentration and DE values after saccharification were less than ± 0.15% (w/v) and ± 0.35%, respectively. Correlation coefficients for model predictions and experimental values were 0.60 and 0.91 for liquefaction and 0.84 and 0.71 for saccharification of amylose and amylopectin, respectively. Model predictions for glucose (R2 = 0.69-0.79) and DP4+ (R2 = 0.8-0.68) were more accurate than the maltotriose and maltose for hydrolysis of high-amylose and waxy corn starch. For yellow dent corn, simulation predictions for glucose were accurate (R2 > 0.73) indicating that the model can be used to predict the glucose concentrations during starch hydrolysis.

Ultrafiltration of thin stillage from conventional and e-mill dry grind processes.

We used ultrafiltration (UF) to evaluate membrane filtration characteristics of thin stillage and determine solids and nutrient compositions of filtered streams. To obtain thin stillage, corn was fermented using laboratory methods. UF experiments were conducted in batch mode under constant temperature and flow rate conditions. Two regenerated cellulose membranes (10 and 100 kDa molecular weight cutoffs) were evaluated with the objective of retaining solids as well as maximizing permeate flux. Optimum pressures for 10 and 100 kDa membranes were 207 and 69 kPa, respectively. Total solids, ash, and neutral detergent fiber contents of input TS streams of dry grind and E-Mill processes were similar; however, fat and protein contents were different (p < 0.05). Retentate obtained from conventional thin stillage fractionation had higher mean total solids contents (27.6% to 27.8%) compared to E-Mill (22.2% to 23.4%). Total solids in retentate streams were found similar to those from commercial evaporators used in industry (25% to 35% total solids). Fat contents of retentate streams ranged from 16.3% to 17.5% for the conventional process. A 2% increment in fat concentration was observed in the E-Mill retentate stream. Thin stillage ash content was reduced 60% in retentate streams.