Developing a computer-controlled simulated digestion system to ...

Published November 24, 2014

Developing a computer-controlled simulated digestion system to predict the concentration of metabolizable energy of feedstuffs for rooster1 F. Zhao,*2,3 L. Q. Ren,* B. M. Mi,* H. Z. Tan,† J. T. Zhao,† H. Li,* H. F. Zhang,* and Z. Y. Zhang* *The State Key Laboratory of Animal Nutrition, Institute of Animal Sciences, Chinese Academy of Agricultural Sciences, Beijing 100193, China; and †Wen’s Food Group Co. Ltd., Guangdong 527439, China

ABSTRACT: Four experiments were conducted to evaluate the effectiveness of a computer-controlled simulated digestion system (CCSDS) for predicting apparent metabolizable energy (AME) and true metabolizable energy (TME) using in vitro digestible energy (IVDE) content of feeds for roosters. In Exp. 1, the repeatability of the IVDE assay was tested in corn, wheat, rapeseed meal, and cottonseed meal with 3 assays of each sample and each with 5 replicates of the same sample. In Exp. 2, the additivity of IVDE concentration in corn, soybean meal, and cottonseed meal was tested by comparing determined IVDE values of the complete diet with values predicted from measurements on individual ingredients. In Exp. 3, linear models to predict AME and TME based on IVDE were developed with 16 calibration samples. In Exp. 4, the accuracy of prediction models was tested by the differences between predicted and determined values for AME or TME of 6 ingredients and 4 diets. In Exp. 1, the mean CV of IVDE was 0.88% (range = 0.20 to 2.14%) for corn, wheat, rapeseed meal, and cottonseed meal. No difference in IVDE was observed between 3 assays of an ingredient, indicating that the IVDE assay is repeatable under these

conditions. In Exp. 2, minimal differences ( 0.96) between AME or TME and IVDE concentrations. In fact, our results indicate that less than 4% of the observed variation in the AME or TME concentration of calibration samples results from factors other than IVDE. The RSD can be used to evaluate the accuracy of prediction models. The accuracy of predicting AME or TME from IVDE may be affected by the range of AME or TME in calibration samples because the degree of variation in the AME or TME directly corresponds to the RSD (Carré, 1990). In the present study, the RSD for AME (146 kcal/kg of DM) and TME (148 kcal/kg of DM) prediction models were similar to the RSD reported for an AME prediction model based on the IVDE of 11 rooster diets (145 kcal/kg of DM; Clunies et al., 1984). However, the range of AME of our 16 calibration samples exceeded that of the 11 rooster diets (1,365 to 3,540 vs. 2,398 to

IVDE, kcal/kg Ingredient Corn B Soybean meal B Cottonseed meal A Diet CSM CCM CSCM

Determined

Calculated3 SEM

P-value

Deviation,4 %

3,806 ± 13 3,301 ± 14

– –

– –

– –

– –

2,701 ± 26

–

–

–

–

3,716 ± 18 3,590 ± 10 3,644 ± 7

3,695 3,580 3,633

8 4 3

0.0626 0.0855 0.0539

0.57 0.28 0.30

1Values are mean of 5 replicates per sample and expressed as mean ± SD. 2CSM = corn–soybean meal diet; CCM = corn–cottonseed meal diet; CSCM = corn–soybean meal–cottonseed meal diet. 3Values

are calculated according to the determined IVDE of individual ingredients and their concentrations in the diet. 4Values are calculated as (determined – calculated)/determined ×100%.

3,356 kcal/kg of DM), resulting in greater accuracy of our prediction model compared to that of Clunies et al. (1984). The RSD in the present study was less than that reported for an AME prediction model based on IVDE of 71 rooster diets (RSD = 152 kcal/kg of DM), with AME ranging from 2,120 to 3,339 kcal/kg of DM (Valdes and Leeson, 1992). The present RSD was substantially less than that reported for predicting the AME of 9 feed ingredients, including grains, soybean meals and grain processed byproduct (RSD = 380 kcal/kg of DM), with AME ranging from 2,491 to 3,770 kcal/kg of DM (Valdes and Leeson, 1992). These results indicate that the accuracy of AME and TME prediction models in the present study exceed those of conventional in vitro digestion procedures reported by Clunies et al. (1984) and Valdes and Leeson (1992). Our RSD was substantially less than those found with the popular in vitro digestion procedure developed by Boisen and Fernández (1997) to predict the AME of 94 starchy grains and cereal byproducts (RSD = 265 kcal/kg of DM) or 52 oil seeds or oil seed byproducts (RSD = 379 kcal/kg of DM) for roosters (Losada et al., 2009, 2010). Therefore, we conclude that the accuracy of AME prediction model developed with CCSDS exceeds that of the manual in vitro digestion procedures reported by Boisen and Fernández (1997). To further evaluate the accuracy of our models, the predicted values were compared with the determined AME and TME values. The mean difference among repeated assays of the same feed was 185 kcal/kg of DM for AME determined with in vivo bioassay (Bourdillon et al., 1990) or 3% of relative errors for TME (Dale and Fuller, 1980, 1986). A difference between determined and predicted

1544

Zhao et al.

Table 5. The determined and predicted values for apparent metabolizable energy (AME) and true metabolizable energy (TME) in 16 calibration samples (DM basis)1 AME, kcal/kg Category Grain

Plant protein

Grain byproduct Other ingredient

Ingredient

IVDE,2 kcal/kg

Determined3

Predicted4

Corn A Corn B Corn C Wheat A Soybean meal A Soybean meal B Cottonseed meal A Cottonseed meal B Rapeseed meal A Rapeseed meal B Sunflower meal DDGS6 Cassava Coconut meal Palm meal Sesame meal

3,718 ± 9 3,806 ± 13 3,714 ± 7 3,621 ± 18 3,205 ± 19 3,301 ± 14 2,701 ± 26 2,042 ± 26 2,031 ± 23 2,547 ± 49 2,866 ± 13 3,154 ± 16 2,258 ± 23 2,162 ± 15 1,618 ± 19 1,993 ± 11 2,796 1,618 3,806

3,456 ± 57 3,540 ± 39 3,448 ± 54 3,255 ± 94 2,851 ± 156 2,904 ± 102 2,428 ± 122 1,374 ± 181 1,527 ± 94 1,981 ± 173 2,584 ± 110 2,822 ± 87 2,116 ± 43 1,948 ± 132 1,365 ± 101 1,420 ± 110 2,439 1,365 3,540

3,419 3,512 3,414 3,316 2,874– 2,975 2,338 1,639 1,627 2,175 2,514 2,820 1,868 1,766 1,188 1,587 2,439 1,188 3,512

Mean Minimum Maximum RSD7

146

TME, kcal/kg Difference

Determined3

Predicted5

Difference

37 28 34 –61 –23 –71 90 –265 –100 –194 70 2 248 182 177 –167

3,940 ± 58 4,024 ± 36 3,951 ± 55 3,831 ± 92 3,341 ± 154 3,394 ± 101 2,927 ± 122 1,948 ± 181 2,020 ± 95 2,527 ± 172 3,053 ± 110 3,313 ± 86 2,692 ± 43 2,505 ± 132 1,850 ± 100 1,910 ± 110 2,952 1,850 4,024

3,888 3,980 3,884 3,786 3,349 3,449 2,820 2,128 2,117 2,658 2,993 3,296 2,355 2,254 1,683 2,077 2,920 1,683 3,980

52 44 67 45 –8 –55 107 –180 –97 –131 60 17 337 251 167 –167

148

1Values are mean of 5 replicates per sample and expressed as mean ± SD. 2IVDE = in vitro digestible energy. 3Determined with 24 roosters for each sample and expressed as mean ± SD. 4Values are calculated based on AME = 1.062 × IVDE – 530 (R2 = 0.97, P < 0.001). 5Values are calculated based on TME = 1.050 × IVDE – 16 (R2 = 0.97, P < 0.001). 6DDGS = dried distiller’s grains with solubles. 7RSD = residual standard deviation.

Figure 3. Linear model to predict apparent metabolizable energy (AME) from in vitro digestible energy (IVDE) of 16 calibration samples including 4 grains (3 sources of corn and 1 of wheat), 7 common protein meals (2 sources of soybean meal, 2 sources of cottonseed meal, 2 sources of rapeseed meal, and 1 source of sunflower meal), 1 grain processed product (distiller’s dried grains with solubles), and 4 other ingredients (1 cassava, 1 coconut meal, 1 palm meal, and 1 sesame meal). Apparent metabolizable energy values were determined with 24 roosters for each sample and expressed as mean ± SD, and IVDE values were the mean of 5 replicates per sample. RSD = residual standard deviation.

Figure 4. Linear model to predict true metabolizable energy (TME) from in vitro digestible energy (IVDE) of 16 calibration samples including 4 grains (3 sources of corn and 1 of wheat), 7 common protein meals (2 sources of soybean meal, 2 sources of cottonseed meal, 2 sources of rapeseed meal, and 1 source of sunflower meal), 1 grain processed product (distiller’s dried grains with solubles), and 4 other ingredients (1 cassava, 1 coconut meal, 1 palm meal, and 1 sesame meal). True metabolizable energy values were determined with 24 roosters for each sample and expressed as mean ± SD, and IVDE values were the mean of 5 replicates per sample. RSD = residual standard deviation.

1545

Prediction of metabolizable energy

AME of less than 100 kcal/kg of DM has been considered an acceptable accuracy for published in vitro digestion procedures (Valdes and Leeson, 1992). The difference between determined and predicted values for AME was less than 100 kcal/kg of DM (Table 5) for 10 of the 16 (62.5%) calibration samples; 4 (25%) differences ranged from 100 to 200 kcal/kg of DM and only 2 (12.5%) differences exceeded 200 kcal/kg of DM. In total, differences between determined and predicted AME of were less than 200 kcal/ kg of DM for a substantial portion (87.5%) of calibration samples. In comparison, Valdes and Leeson (1992) reported differences of less than 100 and 200 kcal/kg of DM in 42.2 and 70.4% of diet samples, respectively. Samples with differences less than 100 kcal/kg of DM encompassed various ingredient types (grains, soybean meal, DDGS, sunflower meal, cottonseed meal, and rapeseed meal), indicating that our assay is accurate for a range of ingredients. The difference between determined and predicted TME was less than 100 kcal/kg of DM for 9 of the 16 (56.3%) samples, which included grains, soybean meals, DDGS, sunflower meal, and rapeseed meal, supporting the validity of our unique in vitro assay. Furthermore, differences ranged from 100 to 200 kcal/kg of DM in 5 (31.3%) samples and 2 (12.5%) differences were greater than 200 kcal/kg of DM. Correlations between AME and IVDE or between TME and IVDE are dependent on both the simulation of in vivo digestion and the precision of determination. In the present study, the mean CV of repeatedly determined IVDE, AME, and TME for the same sample were 0.75, 5.03, and 3.93%, respectively (Table 6), indicating greater precision for measuring IVDE than AME and TME. Similar results have been found when measuring in vitro DM digestibility and IVDE of feed for pigs (Regmi et al., 2008). In the present study, 4 grains showed the CV of less than 0.50, 2.89, and 2.40% for IVDE, AME, and TME, respectively, and the differences between determined and predicted values for AME and TME were less than 67 kcal/kg of DM. These results indicate the good accuracy of using CCSDS to predict the AME and TME of grains. However, for plant-based protenaceous ingredients (e.g., DDGS, soybean meal, sunflower meal, cottonseed meal, and rapeseed meal), the CV of less than 1.27, 6.16, and 4.70% were observed for IVDE, AME, and TME, respectively, and differences between determined and predicted values were less than 107 kcal/kg of DM. Cottonseed meal B and rapeseed meal B had exceptionally high CV for AME, TME, and IVDE relative to other samples, and differences between determined and predicted values ranged from 194 to 265 and 131 to 180 for AME and TME, respectively. These results indicate the accuracy of predicted AME or TME is acceptable when the precision of the AME or TME is satisfactory. Therefore, our results indicate that the accuracy of predicted

Table 6. The CV of apparent metabolizable energy (AME), true metabolizable energy (TME), and in vitro digestible energy (IVDE) in of 16 calibration samples CV,1 % Category Grain

Plant protein

Corn byproduct Other ingredient

Mean Minimum Maximum

Ingredient

AME

TME

IVDE

Corn A Corn B Corn C Wheat A Soybean meal A Soybean meal B Cottonseed meal A Cottonseed meal B Rapeseed meal A Rapeseed meal B Sunflower meal DDGS2

1.65 1.10 1.57 2.89 5.47 3.51 4.98 13.17 6.16 8.73 4.26

1.47 0.89 1.39 2.40 4.61 2.98 4.17 9.29 4.70 6.81 3.60

0.24 0.34 0.19 0.50 0.56 0.45 0.96 1.27 1.13 1.92 0.45

3.08 2.03 6.78 7.40 7.75 5.03 1.10 13.17

2.60 1.60 5.27 5.41 5.76 3.93 0.89 9.29

0.51 1.02 0.69 1.17 0.55 0.75 0.19 1.92

Cassava Coconut meal Palm meal Sesame meal

1Calculated

based on 6, 6, and 5 observations per sample for AME, TME, and IVDE, respectively. 2DDGS = dried distiller’s grains with solubles.

AME or TME is questionable with unacceptable precision of AME or TME determination. Interestingly, the precision of IVDE, AME, and TME determination was acceptable for cassava (CV ≤ 2.03%), but the differences between predicted and determined values were 247 and 338 kcal/kg of DM for AME and TME, respectively, implying that the relationship between in vitro and in vivo digestion of cassava is not in consistent with other samples. Validation of Apparent Metabolizable Energy and True Metabolizable Energy Prediction Models To validate the CCSDS, the determined and predicted values for AME and TME of 10 validation samples were obtained by in vivo bioassay and the prediction models were developed in Exp. 3, respectively. Differences between determined and predicted values were less than 100 kcal/kg of DM for 3 and 4 of 6 ingredient samples for AME and TME, respectively, and differences less than 25 kcal/kg of DM were found for all diets (Table 7). The RSD were 110 and 98 kcal/kg for AME and TME, respectively, which were less than the prediction models developed in the Exp. 3. This result indicates the accuracy of predicted AME and TME of validation samples exceeds that of calibration samples. Linear regression of predicted and determined values revealed that the intercepts did not differ from 0 (P = 0.2621 and P = 0.1445 for AME and TME, respectively)

1546

Zhao et al.

Table 7. The determined and predicted values for apparent metabolizable energy (AME) and true metabolizable energy (TME) in 10 validation samples (DM basis)1 AME, kcal/kg Category

Ingredient and diet

Determined1

Grain

Wheat B Sorghum Fishmeal Peanut meal Rice gluten meal A Rice gluten meal B CSM CCM CSCM SSCM

3,204 ± 62 2,888 ± 45 3,599 ± 129 2,592 ± 159 3,458 ± 131 3,601 ± 126 3,415 ± 55 3,273 ± 48 3,349 ± 35 3,327 ± 74 3,271 2,592 3,601

Animal product Plant protein Unconventional ingredient Diet4

Mean Minimum Maximum RSD5 Estimates of regression6 Intercept P-value7 Slope P-value8 R2

TME, kcal/kg

Predicted2

Difference

Determined1

Predicted3

Difference

3,293 2,856 3,684 2,703 3,623 3,803 3,416 3,283 3,338 3,352 3,335 2,703 3,803 110

–89 32 –85 –111 –165 –202 –1 –10 11 –25

3,681 ± 63 3,380 ± 44 4,099 ± 129 3,129 ± 157 3,884 ± 131 4,128 ± 126 3,894 ± 54 3,766 ± 49 3,823 ± 34 3,803 ± 74 3,759 3,129 4,128

3,764 3,331 4,150 3,180 4,091 4,268 3,886 3,754 3,808 3,822 3,805 3,180 4,268 98

–83 49 –51 –51 –207 –140 8 13 15 –19

290 0.262 0.89 0.177

427 0.145 0.88 0.109

0.95

0.95

1Determined with 24 roosters for each sample and expressed as mean ± SD. 2Values are calculated based on AME = 1.062 × in vitro digestible energy (IVDE) – 530 (R2 = 0.97, P < 0.001). 3Values are calculated based on TME = 1.050 × IVDE – 16 (R2 = 0.97, P < 0.001). 4CSM

= corn–soybean meal diet; CCM = corn–cottonseed meal diet; CSCM = corn–soybean meal–cottonseed meal diet; SSCM = starch–soybean meal–cottonseed meal diet. 5RSD = residual standard deviation. 6Regression of determined values on predicted values for AME or TME. 7H : intercept = 0; H : intercept ≠ 0. 0 α 8H : slope = 1; H : slope ≠ 1. 0 α

and slopes did not differ from 1 (P = 0.1766 and P = 0.1094 for AME and TME, respectively; Table 7). This result indicates the observed AME and TME values of 10 validation samples are equivalent to those predicted by our model. Therefore, it can be concluded that there are similar relationships between AME and IVDE and between TME and IVDE in validation samples to these in calibration samples. The CV of IVDE, AME, and TME determination were less than 0.91, 3.58, and 3.15%, respectively, for 7 out of 10 samples, and differences between determined and predicted values were less than 100 kcal/kg of DM for AME and TME (Table 8). Peanut meal had the greatest CV of the 10 validation samples (CV = 1.45, 6.13, and 5.05% for IVDE, AME, and TME, respectively), and differences between determined and predicted values were 111 and 51 kcal/kg of DM for AME and TME, respectively. However, CV were less than 0.43, 3.79, and 3.37% for IVDE, AME, and TME, respectively, for 2 rice gluten meal samples, but differences between predicted and determined values exceeded 140 kcal/kg of DM for

AME and TME, indicating a poor relationship between in vitro and in vivo digestion of rice gluten meal. In conclusion, our results showed that a new in vitro digestion method can successfully predict the energetic value of ingredients and complete diets for roosters. Repeatability and additivity were acceptable for IVDE determined by CCSDS. In the total of 26 samples, the AME and TME values were accurately predicted for a variety of ingredients and complete diets, but AME and TME values were not accurate for some ingredients (e.g., cottonseed meal B, rapeseed meal B, coconut meal, palm meal, sesame meal cassava, and rice gluten meal), indicating further improvements are needed to successfully use our in vitro digestion method for these feed ingredients. Literature cited Adeola, O., D. Ragland, and D. King. 1997. Feeding and excreta collection techniques in metabolizable energy assays for ducks. Poult. Sci. 76:728–732.

Prediction of metabolizable energy

Table 8. The CV of apparent metabolizable energy (AME), true metabolizable energy (TME), and in vitro digestible energy (IVDE) in 10 validation samples1 Category Grain Animal product Plant protein Unconventional ingredient Diet2

Mean Minimum Maximum

CV,1 %

Ingredient and diet

AME

TME

IVDE

Wheat B Sorghum Fishmeal Peanut meal Rice gluten meal A Rice gluten meal B

1.94 1.56 3.58 6.13 3.79 3.50

1.68 1.33 3.15 5.05 3.37 3.08

0.44 0.66 0.91 1.45 0.43 0.25

CSM CCM CSCM SSCM

1.61 1.47 1.05 2.22 2.69 1.05 6.13

1.39 1.30 0.89 1.95 2.32 0.89 5.05

0.48 0.28 0.19 0.63 0.57 0.19 1.45

1Calculated

based on 6, 6, and 5 observations per sample for AME, TME, and IVDE, respectively. 2CSM = corn–soybean meal diet; CCM = corn–cottonseed meal diet; CSCM = corn–soybean meal–cottonseed meal diet; SSCM = starch–soybean meal– cottonseed meal diet.

Alvarenga, R. R., P. B. Rodrigues, M. G. Zangeronimo, R. T. F. Freitas, R. R. Lima, A. G. Bertechini, and E. J. Fassani. 2011. Energetic values of feedstuffs for broilers determined with in vivo assays and prediction equations. Anim. Feed Sci. Technol. 168:257–266. Association of Official Analytical Chemists (AOAC). 1990. Official methods of analysis. 15th ed. AOAC, Arlington, VA. Boisen, S., and B. O. Eggum. 1991. Critical evaluation of in-vitro methods for estimating digestibility in simple-stomach animals. Nutr. Res. Rev. 4:141–162. Boisen, S., and J. A. Fernández. 1997. Prediction of the total tract digestibility of energy in feedstuffs and pig diets by in vitro analyses. Anim. Feed Sci. Technol. 68:277–286. Bourdillon, A., B. Carré, L. Conan, J. Duperray, G. Huyghebaert, B. Leclercq, M. Lessire, J. McNab, and J. Wiseman. 1990. European reference method for the in vivo determination of metabolizable energy with adult cockerels: Reproducibility, effect of food intake and comparison with individual laboratory methods. Br. Poult. Sci. 31:557–565. Carabaño, R., N. Nicodemus, J. García, G. Xiccato, A. Trocino, J. J. Pascual, L. Falcao-e-Cunha, and L. Maertens. 2008. In vitro analysis, an accurate tool to estimate dry matter digestibility in rabbits. Intra- and inter-laboratory variability. World Rabbit Sci. 16:95–203. Carré, B. 1990. Predicting the dietary energy value of poultry feeds. In: J. Wiseman and D. J. A. Cole, editors, Feedstuff evaluation. Butterworths, London, UK. p. 283–300. Clunies, M., S. Leeson, and J. D. Summers. 1984. In vitro estimation of apparent metabolizable energy. Poult. Sci. 63:1033–1039.

1547

Dale, N. M., and H. L. Fuller. 1980. Additivity of true metabolizable energy values as measured with roosters, broiler chicks and poults. Poult. Sci. 59:1941–1942. Dale, N. M., and H. L. Fuller. 1986. Repeatability of true metabolizable energy versus nitrogen corrected true metabolizable energy values. Poult. Sci. 65:352–354. Hong, D., D. Ragland, and O. Adeola. 2001. Additivity and associative effects of metabolizable energy and amino acid digestibility in barley and canola meal for white pekin ducks. Poult. Sci. 80:1600–1606. Kaps, M., and W. Lamberson. 2004. Biostatistics for animal science. CABI Publishing, Wallingford, UK. Losada, B., P. García Rebollar, C. Álvarez, P. Cachaldora, M. A. Ibánez, J. Méndez, and J. C. De Blas. 2010. The prediction of apparent metabolisable energy content of oil seeds and oil seed by-products for poultry from its chemical components, in vitro analysis or near-infrared reflectance spectroscopy. Anim. Feed Sci. Technol. 160:62–72. Losada, B., P. García Rebollar, P. Cachaldora, C. Álvarez, J. Méndez, and J. C. De Blas. 2009. A comparison of the prediction of apparent metabolisable energy content of starchy grains and cereal by-products for poultry from its chemical components, in vitro analysis or near-infrared reflectance spectroscopy. Span. J. Agric. Res. 7:813–823. Muztar, A. J., and S. J. Slinger. 1981. A comparison of the true and apparent metabolizable energy measures using corn and soybean meal samples. Poult. Sci. 60:611–616. NRC. 1994. Nutrient requirements of poultry. 9th rev. ed. Natl. Acad. Press, Washington, DC. Regmi, P. R., W. C. Sauer, and R. T. Zijlstra. 2008. Prediction of in vivo apparent total tract energy digestibility of barley in grower pigs using an in vitro digestibility technique. J. Anim. Sci. 86:2619–2626. Ren, L. Q., H. Z. Tan, F. Zhao, J. T. Zhao, J. Z. Zhang, and H. F. Zhang. 2012b. Using corn starch as basal diet to determine the true metabolizable energy of protein feedstuffs in Chinese Yellow chickens. Poult. Sci. 91:1394–1399. Ren, L. Q., F. Zhao, H. Z. Tan, J. T. Zhao, J. Z. Zhang, and H. F. Zhang. 2012a. Effects of dietary protein source on the digestive enzyme activities and electrolyte composition in the small intestinal fluid of chickens. Poult. Sci. 91:1641–1646. Sakamoto, K., T. Asano, A. Furuya, and S. Takahashi. 1980. Estimation of in vivo digestibility with the laying hen by an in vitro method using the intestinal fluid of the pig. Br. J. Nutr. 43:389– 391. Sauvant, D., J.-M. Perez, and G. Tran, editors. 2004. Tables of composition and nutritional value of feed materials. Wageningen Acad. Publ., Wageningen, the Netherlands. Sibbald, I. R. 1976. A bioassay for true metabolizable energy in feedingstuffs. Poult. Sci. 55:303–308. Sibbald, I. R. 1977. A test of the additivity of true metabolizable energy values of feedingstuffs. Poult. Sci. 56:363–366. Sturkie, P. D. 1976. Avian physiology. 3rd ed. Springer-Verlag, New York. Valdes, E. V., and S. Leeson. 1992. Measurement of metabolizable energy in poultry feeds by an in vitro system. Poult. Sci. 71:1493–1503.