Food & Feed Research

NONLINEAR MODELS FOR PREDICTING METABOLIZABLE ENERGY OF POULTRY DIETS*

JOURNAL No:
Volume 38, Issue 1
PAGES
33-38
KEYWORDS

metabolizable energy, mathematical model, regression analysis, poultry, diets

TOOLS
Čolović R. Radmilo**1, Palić V. Dragan1, Modika Y. Kedibone2, Barnes Penny2
1University of Novi Sad, Institute of Food Technology, 21000 Novi Sad,
Bulevar cara Lazara 1, Serbia
2Agricultural Research Council, Pretoria, Hatfield 0001, Republic of South Africa

*This paper is a result of the research within the project III046012 financed by the Ministry of Science and Technological Development, Republic of Serbia. This study was also supported by the Ministry of Agriculture, Republic of South Africa.


ABSTRACT

ABSTRACT

Determination of energy value of the diets is very important in animal feed industry. The amount of available energy of poultry feed is described either by metabolizable energy (ME) or by organic matter digestibility (OMD). Due to expensive and time-consuming process of in vivo determination of ME, there has been a considerable interest in developing rapid, low-cost and accurate methods for ME determination. The aim of this study was to develop equations for predicting of ME of poultry diets. Twenty one complete diets for poultry were used in this study. Regression analysis was used to generate mathematical models for prediction of response value of true metabolizable energy (TMEn). Independent variables in models were: in vitro digestibility (%, x1), crude protein (%, x2), crude fat (%, x3), crude fibre (%, x4) and ash content (%, x5). The six polynomial equations are proposed in this study. Each equation is describing individual and interaction effects of three factors on the TMEn. In vitro digestibility was taken into account in all equations. The root mean square of most of equations was lower than 4%, showing that proposed equations had very good prediction of experimental results. Equations with crude protein as one of independent variables have low fit of experimental data (R2 < 0,8), indicating that crude protein did not have strong influence on TMEn. In most of equations in vitro digestibility had significant linear, quadratic or interactional effect on TMEn with relatively good fit of data.

Introduction

Achieving appropriate nutritional and physical characteristics of animal feed is very important for animal health and growth. Once the nutrient requirements of the animals were established, a correct balanced diet can be formulated if the accurate nutrient composition of feedstuffs is known (Čolović et al., 2010; Dale and Batal, 2002). Energy value of diets is of great importance for animal feed manufacturers and end users. The amount of available energy in feeds is described either by its metabolizable energy (ME) or by organic matter digestibility (OMD), (Palić and Leeuw, 2009; Pojić et al., 2008). ME determination of diets requires the use of live animals, appropriate sample collection, ME assay trial, and determination of energy content of feed ingredients and collected excreta (Elkin, 1987; Mohamed, 1984).
In vivo determination of ME can be expensive in terms of time and resources, thus it is important to develop rapid, inexpensive, and accurate methods for ME prediction which could be helpful to manufacturers and nutritionists in monitoring of animal feed quality. There has been a considerable and continuous interest to develop equations for prediction of ME (Perai et al., 2010; Robbins and Firman, 2005; Zhang et al., 1994).
The aim of this study was to develop equations for rapid prediction of metabolizable energy of diets for poultry.

MATERIAL AND METHODS

Chemical analysis

Twenty one complete diets for broilers were used in this study. Dry matter, crude protein, crude fat, crude fibre and ash of the diets were determined according to AOAC official methods (AOAC, 2000). A modified method of Boisen and Fernandez (1997) for estimating the enzymatic digestibility of organic matter (EDOM) was used. For determination of in vivo TMEn, a method of Fisher and McNab (1987) was followed.

Experimental design and statistical analysis

Regression analysis was used to generate mathematical models for prediction of response value of true metabolizable energy (TMEn (MJ/Kg DM)). Independent variables in models were: in vitro digestibility (%, x1), crude protein (%, x2), crude fat (%, x3), crude fibre (%, x4) and ash content (%, x5). The response was related to selected variables by secondorder polynomial model. The generalized model proposed for the response is given in equation below:

where Y represents the experimental response, b0, bi, bii, and bij are constants and regression coefficients of the model, and xi and xj are uncoded values of independent variables. STATISTICA software version 9 (Statsoft, Tulsa, OK, USA) was used for regression analysis of experimental data. The software generated regression coefficients for each of the combinations of the independent variables and their significances were determined using the p-values generated through the t‑tests. Adequacy of predicted model was evaluated by coefficient of determination (R2) and magnitude of root mean square (RMS):

 

where Yexp and Ypred are the experimental and predicted values of the response (TMEn (MJ/Kg DM)) and N is the number of experimental values.

Results and Discussion

The chemical composition of complete diets used in this study is shown in Table 1.
The six polynomial equations are proposed in this study. Each equation is describing individual and interaction effects of three factors on the TMEn. In vitro digestibility was taken into account in all equations. Regression equation coefficients are presented in Table 3. Subscript numbers of the coefficients are denoting factors to whom they are related. The significance of the coefficients is evaluated by Student’s t-test and p-values. Bold numbers denote values significant at 95% level.
Equation 1 (in vitro digestibility, crude protein and crude fat) shows that linear and quadratic effect of in vitro digestibility on TMEn is significant (p ≤ 0,05). Equations 2 (in vitro digestibility, crude protein and crude fibre) and 3 (in vitro digestibility, crude protein and crude fibre) are not having significant regression coefficients. All three equations are having relatively low coefficient of determination (R2 < 0,8). Equation 4 (in vitro digestibility, crude fat and crude fibre) indicated that linear and quadratic influence of crude fat and crude fibre was significant, as well as the interaction of all three factors. Obtained coefficient of determination was relatively higher in comparison with equations 1, 2 and 3. Equation 5 (in vitro digestibility, crude fat and ash) is showing significant linear and quadratic influence of in vitro digestibility, with high coefficient of determination (R2 = 0,93). Linear and quadratic influence of in vitro digestibility in equation 6 (in vitro digestibility, crude fibre and ash), likewise in equations 1 and 5, was significant.
Coefficient of determination of equation 6 was R2 = 0,87. The EDOM and TMEn values of complete poultry diets are shown in Table 2.
Table 1.Chemical composition of complete diets.

Diet

Dry matter (%)

Crude protein
(g/100g DM)

Crude fat
(g/100g DM)

Crude fibre
(g/100g DM)

Crude ash
(g/100g DM)

1

89,85

27,37

11,06

3,58

8,29

2

89,59

21,79

7,50

2,98

7,03

3

88,90

16,63

11,06

4,09

6,43

4

89,84

20,78

14,88

3,68

7,35

5

89,57

17,19

9,74

7,25

11,39

6

89,31

21,88

9,56

8,87

7,02

7

85,30

24,72

9,50

2,73

4,28

8

86,35

25,99

5,72

2,78

3,81

9

88,68

25,01

3,17

2,12

5,86

10

88,31

24,61

6,08

2,65

5,42

11

88,77

25,38

3,56

2,49

6,08

12

87,99

24,65

8,82

2,38

5,21

13

88,50

22,35

5,51

2,11

5,22

14

88,43

19,97

7,50

4,24

5,48

15

88,26

21,40

4,32

2,06

5,94

16

88,25

22,66

7,37

2,16

6,06

17

88,28

26,22

2,91

3,05

6,21

18

88,06

26,39

3,21

1,90

5,32

19

88,21

26,53

3,16

2,69

5,63

20

88,16

25,84

2,89

4,07

5,57

21

88,23

26,77

2,79

2,57

2,25

Table 2.Enzymatic degistibility of organic matter (EDOM) and true metabolizable energy (TMEn) of poultry diets.

Diet

EDOM (%)

TMEn (MJ/kg DM)

1

84,01

14,63

2

84,94

14,59

3

83,31

16,92

4

83,78

16,50

5

71,50

12,60

6

75,27

15,28

7

81,60

17,18

8

86,76

15,81

9

81,94

15,28

10

80,29

16,46

11

81,47

15,21

12

81,83

16,51

13

80,75

16,33

14

78,61

15,67

15

79,20

15,99

16

82,68

16,19

17

77,85

14,71

18

79,65

15,46

19

78,60

15,38

20

78,30

14,92

21

78,57

15,46

DM – dry matter

Table 3.Regression coefficients for TMEn of complete diets.

Eq 1

Eq 2

Eq 3

Eq 4

Eq 5

Eq 6

b0

-348,210

b0

-204,868

b0

-189,923

b0

297,747

b0

-330,648

b0

-645,942

Linear

b1

8,848

b1

5,453

b1

5,177

b1

-6,477

b1

8,274

b1

16,183

b2

-0,301

b2

0,473

b2

-0,349

b3

15,586

b3

0,467

b4

-14,945

b3

2,249

b4

-4,944

b3

0,368

b4

-39,663

b5

3,845

b5

7,693

Quadratic

b11

-0,058

b11

-0,036

b11

-0,032

b11

0,037

b11

-0,049

b11

-0,098

b22

-0,029

b22

-0,019

b22

-0,001

b33

0,055

b33

0,004

b44

0,165

b33

-0,039

b44

0,081

b55

0,007

b44

0,416

b55

3,845

b55

0,004

Interaction

b12

0,027

b12

0,007

b12

0,004

b13

-0,188

b13

-0,003

b14

0,165

b13

0,003

b14

0,078

b15

-0,007

b14

0,478

b15

-0,052

b15

-0,109

b23

-0,074

b24

-0,082

b25

-0,011

b34

-0,278

b35

-0,016

b45

0,240

R2

0,77

R2

0,68

R2

0,79

R2

0,86

R2

0,93

R2

0,87

RMS

2,98

RMS

3,45

RMS

2,80

RMS

2,22

RMS

266,36

RMS

2,10

Eq – equation

The RMS of equations 1, 2, 3, 4 and 6 was lower than 4%, showing that those equations had very good prediction of experimental results. The RMS value of equation 5 was very high due to big discrepancy between experimental and predicted results.
Equations with crude protein as one of independent variables had low fit of experimental data, indicating that crude protein did not have strong influence on TMEn. On the other hand, in most of equations in vitro digestibility had significant linear, quadratic or interactional effect on TMEn with relatively good fit of data.

Conclusions

Each of six proposed equations describe individual and interactional effects of three independent variables, and was evaluated by several quality criteria. This could be helpful in choosing appropriate model depending on chosen variable, or quality criteria. The root mean square of most of equations was very low, showing good prediction of experimental data. The polynomial equations proposed in this study could be successfully used for accurate and rapid prediction of TMEn.

АCKNOWLEDGEMENTS

This paper is a result of the research within the project III046012 “Istraživanje savremenih biotehnoloških postupaka u proizvodnji hrane za životinje u cilju povećanja konkurentnosti, kvaliteta i bezbednosti hrane za životinje (Study of modern biotechnological methods in the production of animal feed in order to increase competitiveness, quality and safety of the feed)”, financed by the Ministry of Science and Technological Development, Republic of Serbia. This study was also supported by the Ministry of Agriculture, Republic of South Africa.

REFERENCES

REFERENCES

  1. AOAC (2000). Official Methods of Analysis, 17th Edition, AOAC INTERNATIONAL, Gaithersburg, MD 20877-2417, USA.
  2. Boisen, S., Fernandez, J.A. (1997). Pre-diction of the total tract digestibility of energy in feedstuffs and feed diets by in vitro analyses. Animal Feed Science and Technology, 68, 277-286.
  3. Čolović, R., Vukmirović, Đ., Matulaitis, R., Bliznikas, S., Uchockis, V., Juškiene, V., Lević, J. (2010). Effect of die channel press way length on physical quality of pelleted cattle feed, Food and Feed Research, 37 (1), 1-6.
  4. Dale, N., Batal, A. (2002). Feedstuffs In-gredient Analysis Table. Miller Publishing Co., Minnetonka, MN.
  5. Elkin, R. G. (1987). A review of duck nutrition research, World’s Poult. Sci. J., 43, 84–106.
  6. Fisher, C., McNab, J. M. (1987). Techni-ques for determining the ME content of poultry feeds. In: W. Haresign and D. J. A. Cole (Eds.) Recent Advances in Animal Nutrition, Butterworths, London, 3-17.
  7. Mohamed, K., Leclercq, B., Anwar, A., El-Alaily, H., Soliman, H. (1984). A compa-rative study of metabolizable energy in du-cklings and domestic chicks, Anim. Feed Sci. Technol. 11, 199–209.
  8. Palić, D., Leeuw, K. J. (2009). Comparison of three in vitro methods for determining and predicting the organic matter diges-tibility of complete diets for ruminants, Acta Periodica Technologica, 40, 79-86.
  9. Perai, A. H., Moghaddam, H. N., Asadpour, S., Bahrampour, J., Mansoori, Gh. (2010). A comparison of artificial neural networks with other statistical approaches for the prediction of true metabolizable energy of meat and bone meal, Poultry Science, 89 (7), 1562-1568.
  10. Pojić, M., Palić, D., Mastilović, J., Janić-Hajnal, E. (2008). The introduction of a me-thod for determination of organic matter digestibility in feeds into routine laboratory practice, Food processing, quality and sa-fety, 35 (3), 151-154.
  11. Robbins, D. H., Firman, J. D. (2005). Eva-luation of metabolizable energy of meat and bone meal for chickens and turkeys by various methods, International Journal of Poultry Science, 4 (9), 633-638.
  12. Zhang, W. J., Campbell, L. D., Stothers, S. C. (1994). An investigation of the feasibility of predicting nitrogen-corrected true meta-bolizable energy (TMEn) content in barley from chemical composition and physical characteristics, Canadian journal of Animal Science, 74 (2), 355-360.
  13. AOAC (2000). Official Methods of Analysis, 17th Edition, AOAC INTERNATIONAL, Gaithersburg, MD 20877-2417, USA.






INSTITUTE
OF FOOD
TECHNOLOGY
IN NOVI SAD
www.fins.uns.ac.rs