Prediction models established for net energy and standardized ileal digestible amino acids in regionally sourced fermented soybean meal for growing pigs

Collection of fermented soybean meal samples

Ten FSBM samples were collected from the major produced provinces across China, representing the following regions: Guangdong (sample 1), Sichuan (samples 2 and 10), Hubei (sample 3), Jiangsu (samples 4 and 5), Shandong (samples 6 and 7), Hebei (sample 8), and Tianjin (sample 9). All FSBM samples were prepared by the strain fermentation method and maintained under dry conditions. Owing to commercial confidentiality, detailed microbiological data for the FSBM samples could not be obtained. A total of 10 samples were initially collected and utilized in feeding trials, however, subsequent analysis identified that sample 10 was a distinct protein ingredient. This sample was consequently excluded, leaving 9 FSBM samples for the final analysis (Table 1).

Table 1

Analyzed nutrient composition and source of fermented soybean meal (FSBM) (as-fed basis)

Exp. 1: Available energy experiment

Experimental animals and diets: Twenty-four growing barrows (Duroc×Landrace×Yorkshire) with an initial body weight (BW) of 35.3±3.2 kg were used in this experiment. The experimental diets included a corn-based basal diet and ten test diets formulated by replacing 27.49% of the corn with different FSBM samples. The available energy value of FSBM was calculated using the difference method while maintaining constant levels of dicalcium phosphate, limestone, salt, and the vitamin-mineral premix (Table 2).

Table 2

Ingredient composition and analyzed nutrient concentrations of the experimental diets in Exp. 1

Experimental design and procedure: In this experiment, twenty-four pigs were randomly assigned to two replicated 6×6 Latin square design, with 12 pigs in each square. The pigs in each square were divided into two groups of six, and the two groups entered six similar open-circuit respiration chambers in a crossover manner over six periods. This design provided six replicates per treatment within each square. Each square consisted of six experimental diets: one corn-based basal diet (common to both squares) and five test diets containing different FSBM samples. Thus, while the two Latin squares collectively involved a total of 12 diets, the shared use of the same basal diet across both squares resulted in a total of 11 experimental diets. Throughout each period, all six diets within a square were simultaneously evaluated, with each diet randomly assigned to one respiration chamber.

Each period consisted of 10 days: 5 days for dietary adaptation, 1 day for chambers adaptation, 3 days for feces/urine collection and gas measurement (recording daily O₂ consumption along with CO₂ and CH₄ production), followed by the last day when fasting heat production (FHP) was determined during the last 8 hours from 2230 (day 9) to 0630 (day 10). Pigs were fed one of five experimental diets formulated at 4% of BW, provided as equal-sized meals twice daily at 0830 and 1530, with feed amounts adjusted based on BW measured on days 0, 5, and 10.

Sample collection: During the collection period (days 7 to 9), residual and spilled feed were collected, dried, and weighed daily. Feces were collected each morning at 0830 when the chamber door was opened and immediately stored at −20°C. Urine was collected concurrently into plastic buckets containing 50 mL of 6 N HCl. Then, it was filtered through cotton gauze, and the total volume was recorded while a 5% aliquot was preserved at −20°C. For FHP determination (days 9 to 10), additional 24 h fasting-state urine collections were conducted. At the end of the collection period, all samples were thawed and homogenized. All measurements were carried out according to the methods described by Zhang et al [16] and Noblet [17].

Chemical analyses: The FSBM samples, basal diet, and 10 test diets used in Exp. 1 were analyzed for dry matter (DM, procedure 930.15; AOAC International [18]), crude protein (CP = nitrogen×6.25, procedure 984.13; AOAC International [18]), ether extract (EE; Lyu et al [19]) and ash (procedure 923.03; AOAC International [18]). The neutral detergent fiber (NDF) and acid detergent fiber (ADF) were determined using a fiber analyzer (ANKOM Technology) according to a modification of the procedures of Van Soest et al [20]. The gross energy (GE) of all samples were determined using an adiabatic oxygen bomb calorimeter (Parr Instruments).

Calculations: The apparent total tract digestibility (ATTD) of the experimental diets was calculated according to the methods of Noblet et al [21].

where F_i and F_f are the total intake and fecal output of energy (MJ) or nutrients (gram), respectively, during the collection period.

The DM intake (DMI) from d 7 to 9 in each period was determined by multiplying feed intake by the DM content of diets. The DE of each experimental diet was calculated by subtracting fecal energy from GE. The ME was derived by subtracting urinary energy (UE) and methane energy (CH₄E) from DE. The NE values of each experimental diet were calculated based on gas data: concentrations of O₂, CO₂, and CH₄ were averaged separately for the fed period and the last 8 h of fasting. Respiratory quotient was derived as the ratio of CO₂ production to O₂ consumption (L/d) [22]. Total heat production (THP) and FHP were calculated using established equations [22]. The RE as protein (RE_P) was calculated from nitrogen retention (nitrogen×6.25×5.70, kcal/g), whereas the amount of the RE as lipids (RE_L) was calculated as the difference between RE and RE_P. All methodologies followed Li et al [11], Lyu et al [13], and Adeola [23]. The main calculations are presented as follows:

The difference method was used to calculate the available energy values and ATTD contributions of FSBM, assuming that the minerals and vitamins in the experimental diets were sources of nutrients but not sources of energy [23].

where the energy value_{test diet} and energy value_{basal diet} are the available energy values of the test and basal diets, respectively; r₀ = the corn fraction in the basal diet, r₁ = the corn proportion in the test diet, and r₂ = the FSBM proportion in the test diet. The ATTD_{test diet} and ATTD_{basal diet} values are the ATTD of nutrients in the test and basal diets, respectively.

Statistical analyses: All data for the experiment were analyzed using analysis of variance with the Proc MIXED of SAS (SAS Institute). The LSMEANS statement with Tukey’s adjustment was used to separate mean values. In all analyses, the differences were considered significant if p<0.05. The relationship between NE and chemical composition of FSBM samples was determined using Proc CORR of SAS. The prediction equations for NE of FSBM samples were developed using Proc REG of SAS. The coefficient of determination (R²), root mean square error (RMSE) and Akaike’s information criterion (AIC) were used as the selection criterion for the best fit equations.

Exp. 2: Amino acid digestibility experiment

Animals, diets and experimental design: In this experiment, twenty-two growing barrows (Duroc×Landrace×Yorkshire; initial BW = 48.8±2.8 kg), each fitted with a T-cannula near the distal ileum, were used in a 3-period crossover design to evaluate 11 experimental diets. The experimental diets included one nitrogen-free diet and 10 FSBM test diets. The test diets contained 40% of one of the nine FSBMs as the sole source of protein, while the nitrogen-free diet was formulated according to the methods described by Stein et al [24] to estimate BEL for subsequent determination of SID for CP and AA (Table 3). The surgical implantation of T-cannulas near the distal ileum was carried out following the established procedures described by Stein et al [25]. Each period lasted 7 days. During this period, there were 5 days for dietary adaptation, followed by 2 days for ileal digesta collection from 0800 to 1700. All collection procedures adhered to the description provided by Stein et al [25].

Table 3

Ingredient composition and analyzed nutrient concentrations of the experimental diets in Exp. 2

Sample preparation and chemical analyses: Chromic oxide (Cr₂O₃, 0.3%) was added as an indigestible marker for AA digestibility calculations. Diet and ileal digesta samples from Exp. 2 were hydrolyzed with 6 N HCl (110°C, 24 h) and analyzed for 15 AA using an Amino Acid Analyzer (Hitachi L-8900) [18]. For the sulfur-containing AA, Met and Cys were pre-oxidation with performic acid prior to hydrolysis with 7.5 N HCl under the same conditions, and then analyzed using a Hitachi L-8800 analyzer. Moreover, Trp was determined by High Performance Liquid Chromatography (Agilent 1200 Series; Agilent Technologies) after LiOH hydrolysis (110°C, 22 h). The Cr concentration in samples was determined by Polarized Zeeman Atomic Absorption Spectrometer (Hitachi Z2000) following nitric–perchloric acid wet-ashing.

Calculations: The AID of AA in the diets containing FSBM was calculated according to the following equation [24]:

where AID = the apparent ileal digestibility of an AA or CP (%), AA_digesta = the AA concentration in the ileal digesta (g/kg of DM), AA_diet = the AA concentration in the diets (g/kg of DM), Cr_diet = the chromium concentration in the diet (g/kg of DM), and Cr_digesta = the chromium concentration in the ileal digesta (g/kg of DM).

where BEL = the basal endogenous losses of an AA or CP (g/kg of DM intake), and SID = the standardized ileal digestibility of an AA (%).

Statistical analyses: The statistical analysis for Exp. 2 followed the same methods as Exp. 1. Data for AID and SID were analyzed using Proc MIXED in SAS with Tukey’s test for mean separation. Relationships between digestibility values and chemical composition were assessed using Proc CORR, and prediction equations for SID were developed using Proc REG, with R², RMSE, and AIC as model selection criteria.

Chemical analysis

The chemical composition of FSBM varied among samples, with the coefficient of variation (CV) greater than 10 % for EE, crude fiber (CF), NDF and ADF (Table 1). The CP content of the nine FSBM samples averaged 49.82%, and ranged from 48.02% to 51.10%. The CV values of the 18 types of AA were all below 10%. The average values of Lys and Met were 2.84% and 0.74%, respectively.

Exp. 1: Available energy experiment

Nutrients digestibility and nitrogen balance for diets: The ATTD of DM and organic matter (OM) in FSBM 4 diet was higher than that of FSBM 8 diet (p<0.05), whereas both FSBM 4 and FSBM 9 diets exhibited greater ATTD of GE than the FSBM 8 diet (Table 4). For ATTD of CP, all FSBM diets except FSBM 8 diet were superior to the basal diet (p<0.05). The addition of FSBM increased nitrogen intake and nitrogen retention compared to the basal diet (p<0.05). Although fecal output was highest in FSBM 8 diet (10.57 g/d), FSBM 1, 6 and 8 diets showed higher fecal output than the basal diet (p<0.05).

Table 4

Effect of experimental diets containing different fermented soybean meal (FSBM) sources on nutrient and energy digestibility and nitrogen balance in growing pigs

Energy balance and energy value for experimental diets: The ME intake and FHP were not affected by dietary treatments (Table 5). Pigs fed FSBM 1, 4, 6, and 9 diets showed higher THP than those fed the basal diet, whereas the FSBM 8 diet resulted in lower THP than FSBM 4 diet (p<0.05). For RE_P, all FSBM diets showed higher values compared to the basal diet (p<0.05). The NE-to-ME ratio in FSBM 3 diet, FSBM 4 diet, and FSBM 5 diet was lower than that of basal diet (p<0.05). However, no differences were observed in the ME-to-DE ratio and UE-to-DE ratio.

Table 5

Effect of experimental diets containing different fermented soybean meal (FSBM) sources on energy balance of growing pigs

Nutrient digestibility and energy contents for ingredients: No differences were observed in the energy utilization among the FSBM samples (Table 6). Consistent with the dietary results, the ATTD of DM and OM was higher in FSBM 4 than in FSBM 8 (p<0.05). With the exception of FSBM 6, all other FSBM samples exhibited a higher ATTD of CP compared to FSBM 8 (p<0.05). Furthermore, the ATTD of ADF was higher in FSBM 3, 4, 5, and 9 than in FSBM 8 (p<0.05). The FSBM 4 and 9 showed higher DE value than that of FSBM 1, 6, and 8 (p<0.05). The DE, ME, and NE values of FSBM samples ranged from 15.50 to 18.44 MJ/kg DM, 13.98 to 16.72 MJ/kg DM, and 10.10 to 11.05 MJ/kg DM, respectively. The average NE value of the FSBM ingredient was 10.65 MJ/kg DM.

Table 6

The apparent total tract digestibility, energy utilization, and energy values of the fermented soybean meal (FSBM) samples

Exp. 2: Amino acid digestibility experiment

Apparent ileal digestibility and basal endogenous losses in growing pigs: There were differences in the AID values of Lys, Met, and Trp among the indispensable AA, as well as in those of Asp and Tyr among the dispensable AA (Table 7). The AID of Trp and Asp was higher in FSBM 9 than that of FSBM 1 and 6 (p<0.05). The AID of Tyr in FSBM 9 was higher than in FSBM 6 (p<0.05). The BEL values for indispensable AA ranged from 0.16 g/kg DMI for Trp to 0.79 g/kg DMI for Leu. For dispensable AA, the BEL ranged from 0.05 g/kg DMI for both Ser and Tyr to 0.27 g/kg DMI for Glu.

Table 7

The apparent ileal digestibility (AID) and the averaged basal endogenous losses (BEL, g/kg dry matter intake) of crude protein and amino acids from different fermented soybean meal (FSBM) in Exp. 2

Standardized ileal digestibility: Statistical analysis revealed that the SID values of AA demonstrated variation among FSBM samples (p<0.05) for most AA, with the exception of Arg and Lys among indispensable AA, and Glu, Ser, and Tyr among dispensable AA (Table 8). The SID of Met was highest in FSBM 9, surpassing that of FSBM 3, 4, 5, 6, and 8 (p<0.05). Similarly, the SID values for Iso, Phe, and Trp in FSBM 9 were also greater compared to FSBM 1 and 6 (p<0.05).

Table 8

The standardized ileal digestibility values of crude protein and amino acids from different fermented soybean meal (FSBM) in Exp. 2

Correlation analysis and prediction equations for standardized ileal digestible amino acids and net energy in fermented soybean meal samples: In terms of energy values, the DM content exhibited positive correlations with both GE (r = 0.75, p< 0.05) and NE (r = 0.78, p<0.05; Table 9). Regarding the SID of AA, the ash showed a negative correlation with SID_Lys (r = −0.70, p<0.05), whereas the NDF had a positive correlation with SID_Lys (r = 0.69, p<0.05).

Table 9

Correlation coefficients between chemical composition, energy values, and standardized ileal digestibility of the nine fermented soybean meal samples

The regression equations for NE in the current experiment were developed based on chemical composition and GE (Table 10). Three robust models emerged: (1) an NDF-GE model (R² = 0.81, p<0.01), (2) an EE-GE model (R² = 0.85, p<0.01), and (3) a comprehensive model incorporating EE, NDF and GE (R² = 0.92, p<0.01). The best-fit predictive equations for SID_CP and SID_Lys were as follows: SID_CP = 117.42–0.58×DM–0.40×NDF+2.99×ADF (R² = 0.71, p = 0.079) and SID_Lys = 196.63+1.56×CF+6.08×ADF–24.56×Ash (R² = 0.83, p = 0.022), respectively.

Table 10

Stepwise regression equations to estimate net energy (MJ/kg DM), standardized ileal digestibility of crude protein (SID_CP) and lysine (SID_Lys, %) from nutrient concentrates of fermented soybean meal (DM basis)

Chemical composition of ingredient

The fermentation process enhances the nutritional and functional qualities of SBM. It not only eliminates ANF, reducing their negative impact on digestion, but also enables microbial proteases to break down soybean protein into highly absorbable peptides and free AA, greatly improving digestibility and utilization [2].

The nutrient composition (such as EE, CF, NDF, and ADF) of the FSBM samples exhibited notable variability, likely due to the differences in the sources or strains of SBM used in the fermentation process [26]. Variability in fiber content may stem from inherent differences in soybean composition, inconsistent removal of seed coats, and varying microbial degradation efficiency during fermentation. The average EE content observed was lower than that reported by NRC [27] (1.16% vs. 2.30%), possibly reflecting improvements in modern oil extraction technology. The Lys concentration was consistent with values reported by Yan et al [28] and Huang et al [15], but lower than that reported by Jang and Kim [29]. Variations in Lys content across studies may be attributed to differences in fermentation processes, microbial strains, and analytical methods.

Nutrient digestibility and nitrogen balance for experimental diets and ingredients

The lower ATTD of DM, OM, GE, and CP observed in the FSBM 8 diet, compared to the FSBM 4 diet, can be attributed to differences in nutrient compositions. The higher fiber content (CF, NDF, and ADF) likely physically hindered enzymatic access to nutrients, thereby reducing overall digestibility. Consistent with this finding, the ATTD values of the FSBM samples calculated using the difference method showed similar trends: the FSBM 8 sample exhibited lower ATTD of GE, CP, ADF, and OM than the FSBM 4 sample.

Compared to the basal diet, all FSBM groups showed higher nitrogen intake (42.3 vs. 15.2 g/d) and retention (21.0 vs. 4.2 g/d), consistent with the positive correlation between CP levels and nitrogen utilization. Fecal nitrogen output primarily consists of undigested dietary nitrogen, endogenous losses, and microbial protein [30]. The FSBM 8 diet resulted in higher fecal nitrogen output (10.6 vs. 6.4 g/d on average) and lower ATTD of CP (78.0% vs. 86.7% on average) than other FSBM diets. This may be due to: 1) excessive fermentation-induced protein denaturation and Maillard reactions, and 2) specific microbial strains that increased indigestible microbial protein. These results highlight that the nutritional value of FSBM depends not only on CP level, but also on the overall nutrient composition and the effectiveness of the fermentation process.

Energy balance and energy value for experimental diets and ingredients

The RE_P values were consistent with nitrogen retention, indicating that the addition of FSBM increased RE_P. This improvement can be attributed to the enhanced AA balance in the diet, which effectively promotes protein deposition.

The available energy of FSBM samples was calculated using the difference method, which assumes no interaction between the basal diet and test ingredient. However, in reality, such interactions, particularly from CP, influence energy values. Studies indicate that higher dietary CP levels increase whole-body protein turnover and urinary nitrogen excretion, leading to greater heat increment [21]. Therefore, when applying the difference method to determine the energy value of protein-rich ingredients like FSBM, the difference in CP content between the basal diet and test diets is critical for accurately estimating energy value [30].

The average ME value of the FSBM samples (15.63 MJ/kg DM) was lower than the value listed in Nutrient Requirements of Swine in China (16.52 MJ/kg DM) [31]. This discrepancy can be attributed to two main factors: the lower EE content in the FSBM samples used in this study (1.27% vs. 2.30%), and the higher CP level (average 19.77%) in the FSBM diets. As reported by Kim et al [32], when dietary CP exceeds 17.0% in growing pigs, UE excretion increases, which may result in an underestimation of ME and possibly NE of the test ingredient. Among the different FSBM samples, the DE values of FSBM 8 were lower than those of FSBM 4 and 9. This difference may be attributed to the higher fecal output observed in FSBM 8, as fecal output is negatively correlated with DE [17].

Amino acid digestibility

The comprehensive analysis of AID, BEL, and SID in the present study offers valuable insights into the AA availability of the FSBM samples evaluated. Samples such as FSBM 9 exhibited higher digestibility values than FSBM 1 and 6 in Trp and Asp, suggesting that differences in fermentation conditions, possibly related to microbial strains, process parameters, or substrate composition, distinctly influenced the breakdown of protein fractions and ANF.

The comprehensive analysis of AID, BEL, and SID in the present study offers valuable insights into the AA availability of the FSBM samples evaluated. Significant variations were observed among samples; for instance, the AID of Lys was higher in FSBM 2, 4, and 9 than in FSBM 1 (p<0.05), and the AID of Tyr in FSBM 9 was higher than in FSBM 6 (p<0.05). Samples such as FSBM 9 consistently exhibited superior digestibility values, suggesting that differences in fermentation conditions, possibly related to microbial strains, process parameters, or substrate composition, distinctly influenced the breakdown of protein fractions and ANF.

The BEL values obtained in this study were generally lower than those reported in other studies with growing pigs. For instance, the mean BEL of CP determined in the present study was 5.20 g/kg DMI, substantially lower than the values of 10.53 g/kg DMI reported by Jansman et al [33] and 17.10 g/kg DMI reported by Park et al [34]. Similar trends were observed for many AA, with the mean BEL of Val (0.26 vs. 0.41 g/kg DMI), Thr (0.19 vs. 0.51 g/kg DMI), and Pro (0.08 vs. 1.31 g/kg DMI) being lower than those reported by Jansman et al [33]. It is noteworthy that despite these lower mean values, all BEL values for AA obtained in this study fell within the ranges reported in these meta-analyses [33,34]. Given that all these results were based on the nitrogen-free diet, the considerable variations could be attributed to other factors such as animal age, initial BW, dietary composition, and feeding level [29,34,35]. These variations in BEL are critical to acknowledge, as lower BEL values may amplify the impact of incomplete digestion on SID estimates.

After correcting for BEL, the SID values provided a more accurate estimation of the true digestibility of AA from the FSBM samples. The SID of Met (77.5%) in this study was lower than that of NRC [27] (77.5% vs. 88%) and Nutrient Requirement of Swine in China [31] (77.5% vs. 91%), however, consistent with the research (77.5% vs. 75.3%) by Huang et al [15]. Meanwhile, the SID of Lys (74.7%) in this study remains well within the published range for FSBM, with documented values for SID of Lys ranging from 75% to 88% [29] to approximately 77% [29], which further confirms the plausibility of our results. However, existing studies show certain variations in the SID of AA, which may be attributed to differences in processing and fermentation conditions, methodological influences, and matrix effects [36].

Correlations and prediction equations

Our results demonstrated a negative relationship between EE and NE, which may be attributed to the low content of EE in the FSBM samples, limiting its contribution to NE and thus its role in energy prediction. More notably, a positive correlation was observed between NDF and the SID_Lys (r = 0.69, p<0.05), while the relationship between ADF and SID_Lys also showed a positive, though not statistically significant, trend (r = 0.59, p = 0.097). This finding reflects an important characteristic of the fermentation process: higher fiber content (both NDF and ADF) may indicate more extensive fermentation, during which soluble nutrients are metabolized by microorganisms, leading to a relative increase in fiber concentration. Therefore, the positive association with SID_Lys does not imply that ADF directly enhances digestibility; rather, it serves as an indirect marker of stronger microbial activity.

Although this study has established prediction equations for NE, SID_CP, and SID_Lys in FSBM for growing pigs, the relatively low R² values of the models for SID_CP and SID_Lys indicate limited explanatory power. Thus, further research incorporating larger and more diverse datasets is necessary to improve the accuracy and robustness of these predictive equations.

Several limitations of this study should be acknowledged. The absence of data on critical ANF limits interpretation of the variability in digestibility across FSBM samples. Additionally, commercial confidentiality surrounding microbial strains and fermentation parameters hindered further investigation into the causes of product differences. Future research should combine detailed ANF analysis with standardized digestibility assays and promote greater transparency in fermentation protocols to improve reproducibility and enable reliable cross-study comparisons.