亚洲饲料原料的变异和近红外预测饲料原料营养成份的潜在价值

bonderic

NUTRITIONAL VARIATIONS OF MAJOR FEEDSTUFFS USED IN ASIA:NIRS POTENTIAL TO PREDICT THEIR NUTRITIONAL VALUES

C. Gady, M.L. Liu, Y.G. Liu and P.A. Geraert

Adisseo Asia Pacific P/L, Singapore; Adisseo France S.A.S.

Abstract

      Variability of major feed ingredients used in Asia feed industry is found to be substantial. Our analyses of corn (>200 samples) collected from Asian countries revealed that crude protein content averaged 7.95% with CV 6.16%; of soybean meal (SBM, >300 samples) averaged 45.06% with CV 3.26%; contents of crude fibre and ash exhibited even higher variations. Using in vivo determinations, the Apparent Metabolizable Energy (AME) of Asian corns averaged 3,468 kcal/kg with CV 3.4%; SBM 2,271 kcal/kg with CV 8.2%. Magnitude of difference between the highest and lowest AME measurements was 588 kcal/kg for corn and 804 kcal/kg for SBM. Most of the variations among Asian feedstuffs exceeded the worldwide variability. For many feed companies, Near Infrared Reflectance Spectroscopy (NIRS) has been employed mainly to monitor proximate constituents. This paper presents possibility and efficacy to predict Total and Digestible Amino Acid levels (TAA, DAA) as well as AME in main Asian feedstuffs by NIRS. With a few exceptions, our NIRS calibrations may explain 80-97% variability of TAA, DAA and AME, which enables rapid characterization of specific nutritional values batch to batch at low operating cost and confirms its added value for optimizing feed formulation.

Key words: Feedstuff, NIRS, TAA, DAA, AME

Introduction

      Nutritionists aim at formulating diets using feedstuffs of heterogeneous origin to meet requirements of a specific set of nutrients in attempt to ensure adequate performance and maximum return of a particular category of poultry. This has never been an easy task due to countless factors affecting values of the feedstuff used. Some producers often regard corn, soybean meal and other feed ingredients to be more or less consistent in their categories in terms of nutritional value, regardless of where they are grown and how they are processed and stored. Book values were often used in matrix for diet formulation. More advanced nutritionists upgrade their matrix values from time to time through acquiring new testing data or based on field experience.

      Over the past years, analysis of corn and soybean meal samples from all over the world has revealed that these samples have a remarkable degree of variation. This has a major implication on the daily work and efficiency of nutritionists, feed manufacturers and poultry producers.

Materials and Methods

      In order to survey variation of feed ingredients used in Asia, a few hundreds of corn and soybean meal samples were collected from various feed companies and analysed for contents of moisture, crude protein, fats, crude fibre, ash, calcium and phosphorus. In addition, results of protein solubility and level of xanthophylls were also collected.

Results of amino acid digestibility were measured using caecectomized cocks following procedure described by Green et al. (1987). The Apparent Metabolizable Energy contents (AME or AMEn) were determined in 3-wk-old male broilers. Birds were fed with a starter diet till day 11. On day 12, after 4 h starvation the chicks were randomly distributed into 10 experimental groups of 12 in such a way that the groups had the same average weight. These birds were fed until 21 days with pelleted experimental diets and excreta were collected daily for a 72 h balance period (day 20-22). The AMEn was measured using the European reference method with ad libitum feeding and total excreta collection (Bourdillon et al., 1990).

      The NIRS calibrations were developed using such in vivo measurements. Samples used were ground through a 1 mm sieve, and scanned by NIRSystem model 6500 (FOSS Sweden), for reflectance from 400 nm to 2500 nm (every 2 nm) using a large rectangular cup. The partial least squares (PLS) method was used to obtain NIRS equations for all studied parameters. Cross-validations (SECV) were performed to estimate the prediction accuracy of each equation. In addition, the differences between reference values and predicted values were calculated on additional independent samples to confirm the robustness of the equations.

Results and Discussion

      Results of proximate nutrients are summarized in Table 1. Data of corn (>200 samples) displayed content of crude protein 7.95% with CV 6.16%, moisture 13.19% with CV 6.22%; soybean meal (SBM, >300 samples) showed protein averaged 45.06% with CV 3.26%, moisture 12.21% with CV 5.81%. Contents of crude fat and crude fibre in SBM displayed CV >20% for both corn and SBM, confirming a high degree of variation. Calcium content displayed a CV >70% whilst CV of phosphorus level is relatively low. Protein solubility remains rather consistent for both types of samples. Level of xanthophylls in corn exhibited a high degree of variation with CV 32.5%, explaining wide use of added pigments to compensate this variation in the field.   

Table 1. Variation of proximate nutrients in corn and SBM in Asia (feed basis, %)

	Corn				Soybean meal
	Mean	S.D.	C.V.	No.	Mean	S.D.	C.V.	No.
Crude protein	7.95	0.49	6.16	207	45.06	1.47	3.26	344
Moisture	13.19	0.82	6.22	207	12.21	0.71	5.81	344
Crude fats	3.65	0.84	23.01	342	1.64	0.44	26.83	344
Crude fibre	2.60	0.62	23.88	240	4.61	1.05	22.78	344
Crude ash	1.16	0.22	18.97	393	6.38	0.42	6.58	344
Calcium	0.21	0.16	76.19	189	0.44	0.32	72.73	281
Phosphorus	0.27	0.08	29.63	320	0.61	0.05	8.20	292
Protein solubility	67.11	2.55	3.80	267	81.09	3.60	4.44	319
Xanthophyll, ppm	13.27	4.31	32.48	227	-	-	-	-

      Variations of nutritional composition among Asian corns and soybean meals are compared to the corresponding worldwide variations (Table 2). Most of parameters show a large variability with CV up to 5 to 19%. Surprisingly, measured corn AME values exhibited a rather low CV (around 3%). However, the observed range in the results showed large differences between lowest and highest contents. Compared to the worldwide variability, both CV and range measured among Asian feedstuffs are either similar or superior. This confirms that improving knowledge of both nutrient requirements of the animal and available nutrient contents of the ingredients is a key issue in feed formulation (Bushman, 1998). In the case of the Asian ingredient market, difficulties are then emphasized, at least because a large worldwide feedstuffs variability is added to an existing wide variability among Asian feedstuffs.  

Table 2. Comparison of variations between samples from Asia and Worldwide

	CORN					SOYBEAN MEAL
	Worldwide		Asian			Worldwide		Asian
	Mean	CV, %	Mean	CV, %	Range	Mean	CV%	Mean	CV,%	Range
Total Lys. %	0.24	10.9	0.24	8.6	0.05	2.85	4.9	2.84	5.6	0.69
Dig. Lys., %	0.20	10.0	0.22	18.9	0.13	2.52	5.1	2.47	4.9	0.32
Total Met., %	0.17	11.9	0.17	13.3	0.07	0.63	6.3	0.63	6.5	0.23
Dig. Met., %	0.16	12.5	0.16	13.9	0.07	0.57	7.0	0.57	12.1	0.21
AME,kcal/kg DM	3935	2.5	3913	2.8	598	2504	8.0	2504	7.7	910
AME, kcal/kg	3522	2.6	3468	3.4	588	2236	8.8	2271	8.2	804
AMEn, kcal/kg DM	3800	2.6	3773	2.9	576	2324	8.7	2310	8.0	863
AMEn, kcal/kg	3401	3.2	3344	3.5	568	2075	9.7	2099	8.8	799

The potential of NIRS to predict TAA, DAA and AME

      In this context, any quality control system or tool which could offer an answer to the above concerns on variation may offer main advantages such as rapid reactivity for decision making, low cost, accuracy and ability to produce both quantitative and qualitative results. Most research effort has been focused on acquiring equations to predict digestibility, feed intake and energy content through proximate composition. The predictive capacity of these equations seems very limited. For instance, our in vivo evaluations indicate that the protein content (N x 6.25) of soybean meal is not an accurate predictor of either total or digestible lysine content (R2 =0.43 and 0.46, respectively). In addition, time and cost required to obtain precise nutritional characterization such as total amino acids, digestible amino acids, metabolizable energy through in vitro or in vivo tests, make it impossible to adopt such results for diet formulations in routine. For convenience, nutritional average values per feedstuff supported by books are routinely applied, by doing so a large part of the widely existing variability within a given category of feedstuffs is neglected.

      On the other hand, the NIRS calibrations that we developed explain more than 80% of the variation in total and digestible amino acids for cereals and soybean meals. Both the R2 of the models and the cross validation (1-Vr) vary from 0.80 to 0.98 for all amino acids but methionine being slightly lower (respectively 0.79 and 0.77% for cereals and soybean meals). However, despite the low variability and the high digestibility of methionine in cereals, the R2 of the models and cross validation can explain around 80% of the variation. In addition, the methionine analysis by HPLC is linked to a higher degree of error due to oxidation losses during sample preparation. The prediction errors calculated on protein and total amino acids varied from 0.02 to 0.07 against 0.04 to 0.09 for the corresponding digestible amino acid levels. These ratios are comparable with HPLC average error that is close to 0.05. Calculating prediction errors obtained with cross validation, results confirm similar level of accuracy. The prediction errors associated with the NIRS models of total amino acids are lower than those associated with the prediction of digestible amino acids, which may be explained by the fact that the digestible amino acid reference data integrate two sources of variability coming from chemical analysis on both raw material and excreta. However, considering that the errors derived from predicting digestible amino acids have an average value equal to 0.08 for cereals and 0.04 for soybean meal, these NIRS models can be used as an accurate tool for characterizing contents of total and digestible amino acids.  

      Based on the feedstuffs included into the models, the ranges of AMEn values measured in vivo are 3,569-4,041 kcal/kg DM for corn and 2,061-2,806 kcal/kg DM for soybean meals, respectively. The reported variability in energy concentration in feedstuffs as shown in Table 2 also requires an accurate determination of energy value of feedstuffs that can then guide for specific diet formulations. For all feedstuffs studied, the R2 of the models are above 0.80 and can explain 82 and 91% of the variation for corn and soybean meal, respectively. Both prediction errors and those obtained with cross validations are very similar. The NIRS calibrations predicting corn AMEn are associated to a low error of prediction (SEC = 44 kcal/kg DM; SECV = 45 kcal/kg DM). It is also very close to the reproducibility level obtained. Errors associated to the AMEn prediction for soybean meals are slightly higher (SEC = 66 kcal/kg DM; SECV = 69 kcal/kg DM). Such difference may be attributable to the level of total non-starch polysaccharides (NSP) in soybean meal affecting utilization of nutrients and could explain the higher in vivo reproducibility results observed and then the higher degree of error in calibration. Based on the degree of variability prevalent in Asian corn and soybean meals, these NIRS calibrations can be used as an efficient prediction tool.

      Based on the potential of the technology and its capability in characterizing specific and relevant nutritional value of batch to batch of the same ingredients, the NIRS is therefore a method of choice to monitor variability of current feed ingredients used in Asia.

Conclusion

      Our survey demonstrated that major feed ingredients used in feed formulation exhibit high degrees of variation irrespective of nutritional parameter, and variability of Asian feedstuffs is mostly higher than that of the worldwide data, difficulties are obvious for people in the field of animal feeding. It has been demonstrated that NIRS calibrations can be used for prediction of nutritionally relevant parameters, such as total and digestible amino acids and metabolizable energy of corn and soybean meal. By highlighting the portion and quality of the nutritional information often ignored or underestimated by other tools currently available, NIRS confirms its added value in improving accuracy of feed formulation in commercial practice.   

References

Bourdillon, A., Carre, B., Conan, L., Francesch, M., Fuentes, M., Huyghebaert G., Janssen W.M.M.A., Leclercq B., Lessire M., Mc Nab J., Rigoni M. and J. Wiseman. 1990. European reference method for the in vivo determination of metabolisable energy with adult cockerels: reproducibility, effect of food intake and comparison with individual laboratory methods. British Poultry Sc. 31: 567 – 576

Bushman, D.H.1998. Best cost formulation. Feedmix, 6(2): 18-23.

Green, S., Bertrand, S., Duron, M., Maillard, R. 1987. Digestibilities of amino acids in maize, wheat and barley meals determined with intact and caecectomized cockerels. Br. Poult. Sci., 28: 631-641.

qiaoyongniu

good paper, thank you for your sharing!

晚童

牛人，不知道有没有中文版本