Structural and non-structural carbohydrates profiles affect nutrient availability or digestive behavior. Table 1 shows the structural characteristics of carbohydrates: Comparison between the different grains (wheat, corn) and different types of DDGS (wheat DDGS, corn DDGS and blend DDGS (wheat:corn = 70:30)) from bioethanol production, revealed using the DRIFT mid-infrared molecular spectroscopy. Figure 1 shows a typical DRIFT molecular spectrum in corn DDGS in the region ca. 4000–800 cm⁻¹ showed function groups of biopolymers in complex plant system: N-H and O-H stretch, C-H stretch, amide I and II, C=O carbonyl, CHO and cellulosic compounds.

Compared wheat with corn, the results shows that there was no significant differences (P > 0.05) in the A_Cell (structural CHO, spectral peaks region and baseline: ca. 1485–1188) with average 220.1 IR KM units. The A_Cell was associated major hemi- and cellulosic compounds [7,8]. There was no difference (P > 0.05) in the A_1240 (structural CHO, peak area centered at ca.1240 cm⁻¹ with region and baseline: ca. 1292–1198 cm⁻¹) with average 8.5 IR KM units. The A_1240 spectral parameter is associated with cellulosic compounds [7,8]. There were significant differences (P < 0.05) in the A_CHO (total CHO, peaks region and baseline: ca. 1187–950 cm⁻¹) which is total carbohydrate region [7] and the A_928 (non-structural CHO, spectral peak area centered at ca. 928 cm⁻¹ with region and baseline: ca. 952–910 cm⁻¹) which is associated with non-structural carbohydrate. Both the H_1415 (structural CHO, peak height centered at ca.1415 cm⁻¹ with baseline: ca. 1485–1188 cm⁻¹) and H_1370 (structural CHO, peak height at ca.1370 cm⁻¹ with a baseline: ca. 1485–1188 cm⁻¹) are related to cellulosic compounds [8]. There were no differences (P > 0.05) between the wheat and corn (Table 1).

Compared among the co-products (wheat DDGS, corn DDGS and Blend DDGS) (Table 1), there were significant differences (P < 0.05) in A_1240 (cellulosic compounds) and A_CHO (total CHO). Both corn DDGS and Blend DDGS were higher (P < 0.05) in the A_1240 and A_CHO than wheat DDGS. But there were no differences (P > 0.05) between the corn DDGS and blend DDGS. There were no significant differences (P > 0.05) in the peak area intensities of the A_Cell (average 241.3), A_928 (average 2.0) and A_860 (average 7.8), and in the peak height intensities of H_1415 (average 1.5) and H_1370 (average 1.2). These results show that molecular spectral profiles differed between the grains and between the co-products. In general, the co-products from bioethanol processing had higher peak intensities in the structural carbohydrates and lower intensities in the non-structural carbohydrates. So far, no published results could be used to compare with the results from this study. The different structural and non-structural carbohydrate molecular spectral profiles may be highly related to carbohydrate functionality and quality.

Table 2 shows the comparison between the grains and co-products (DDGS) from bioethanol production in the carbohydrate structural characteristics, revealed using the DRIFT mid-infrared molecular spectroscopy. The results show that the grains had significantly lower peak area intensity of A_1240 (8.5 vs. 36.6, P < 0.05), higher A_928 (17.3 vs. 2.0, P < 0.05) and higher A_860 (20.7 vs. 7.6, P < 0.05) than their co-products from bioethanol processing. There were no differences (P > 0.05) in the peak area intensities of A_Cell and A_CHO with average peak area intensities of 226.8 and 508.1 IR KM units, respectively. There were no differences (P > 0.05) in peak height intensities of H_1415 and H_1370 with average peak height intensities of 1.35 and 1.15 IR KM unit, respectively. Again no publications were found in this area.

The results indicate that the co-products and grain differed in carbohydrate structure. Bioethanol processing changes original grain carbohydrate molecular structure profiles. It increased structural carbohydrate profiles and decreased non-structural carbohydrate profiles. It is expected that these structural differences may impact the co-products carbohydrate utilization and availability. The results demonstrate that molecular spectral analytical technique may reveal differences in carbohydrate molecular structure.

Infrared spectra based on similarity with other spectra could be clustered using CLA analysis [5]. In this study, the Ward’s algorithm method was used without any prior parameterization of the spectral data in the four different IR regions:

Region 1 is the mid-infrared region ca. 4000–827 cm⁻¹;

This method helps discriminate in the structural differences between the grain and its co-products. Figures 2–5 show that two classes can be clear distinguished between corn and corn DDGS, but not between wheat and wheat DDGS. These results indicate that molecular structure between the corn and bioethanol co-product (corn DDGS) were different. These results also indicate that different cereal grains have different responses to bioethanol processing and different sensitivity to heating and fermentation.

The second multivariate analysis tested was a PCA analysis which is a statistical data reduction method [5]. In this study, PCA analysis was used to identify the main sources of variation in carbohydrate conformation at four different regions as mentioned before. Figures 2–5 show that corn and corn DDGS can be grouped in separate ellipses (Figures 2–5d) with no overlapping of groups. The first PC can explain 99% of the variation in the four different spectrum data sets

The co-products from bioethanol production—wheat DDGS, blend DDGS (wheat:corn = 70:30), and corn DDGS as well as original feedstock wheat and corn grains were collected. The chemical characteristics, protein and carbohydrate subfraction profiles, energy values [1], rumen and intestinal disappearance from large samples of the DDGS in bioethanol plants [3,4], and modeling nutrient supply were reported [2].

The experiments were carried out at Saskatchewan Structure Sciences Center (SSSC, Saskatoon, Canada). The methodology to prepare samples for molecular spectroscopy study was published before [5]. Each sample from bioethanol production were finely ground (Retsch ZM-1, Brinkmann Instruments (Canada) LTD, Ontario). Diffuse reflectance infrared (KM unit) Fourier transform spectroscopy was performed using a Bio-Rad FTS-40 with a ceramic IR source and MCT detector (Bio-Rad laboratories, Hercules, California, USA). Data was collected using Win-IR software. Spectra were generated from the mid-IR (4000–800 cm⁻¹) portion of the electromagnetic spectrum with 256 co-added scans and a spectral resolution of 4 cm⁻¹.

Molecular spectral analysis was done with OMNIC 7.2 software (Spectra Tech., USA). The carbohydrate related molecular spectral bands included: A_Cell (structural CHO, peaks area region and baseline: ca. 1485–1188 cm⁻¹), A_1240 (structural CHO, peak area centered at ca. 1240 cm⁻¹ with region and baseline: ca. 1292–1198 cm⁻¹), A_CHO (total CHO, peaks region and baseline: ca. 1187–950 cm⁻¹), A_928 (non-structural CHO, peak area centered at ca. 928 cm⁻¹ with region and baseline: ca. 952–910 cm⁻¹), A_860 (non-structural CHO, peak area centered at ca. 860 cm⁻¹ with region and baseline: ca. 880–827 cm⁻¹), H_1415 (structural CHO, peak height centered at ca. 1415 cm⁻¹ with baseline: ca. 1485–1188 cm⁻¹), H_1370 (structural CHO, peak height at ca. 1370 cm⁻¹ with a baseline: ca. 1485–1188 cm⁻¹). The above carbohydrate associated bands were identified according to the published reports [7–9] and discussed in Yu et al. [10].

Statistical analyses were performed using the MIXED procedure of SAS (version 9.1.3). The model used for the analysis was: Y_ij = μ + T_i + e_ij, where, Y_ij is an observation of the dependent variable ij (molecular spectral parameters: A_Cell, A_1240, A_CHO, A_928, A_860, H_1415, H_1370); μ is the population mean for the variable; T_i is the effect of feeds, as a fixed effect, and e_ij was the random error associated with the observation ij. To compare grain and bioethanol co-products, the model used for the analysis is: Y_ij = μ + T_i + e_ij, where, Y_ij is an observation of the dependent variable ij (molecular spectral parameters: A_Cell, A_1240, A_CHO, A_928, A_860, H_1415, H_1370 in IR KM Units); μ is the population mean for the variable; T_i is the effect of feed (i: 1 = grains; 2 = bioethanol co-products) as a fixed effect, and e_ij is the random error associated with the observation ij. The normality check was used Proc Univariate with Normal and Plot options in SAS

For all statistical analyses, significance was declared at P < 0.05 and trends at P ≤ 0.10. Differences among the treatments were evaluated using a multiple comparison test following the Tukey-Karmer method.

Multivariate molecular spectral analyses, principal component analysis (PCA) and hierarchical cluster analysis (CLA), were performed using Statistica software 6.0 (StatSoft Inc, Tulsa, OK, USA) to classify and distinguish between the inherent structures.