Metabolizable protein requirements for maintenance and gain of growing goats

Abstract

A database of 349 treatment mean observations, representing 3404 goats from 73 publications between 1973 and 2003, was used to determine metabolizable protein (MP) requirements for maintenance (MP m ) and growth (MP g ) of goats. Published CP degradation properties of feedstuffs and proportions of dietary ingredients were used to estimate MP intake (MPI, g/day), which was regressed against ADG, with both variables scaled by BW 0.75 . Goats were classified as meat (≥50% Boer; 60 observations), dairy (selected for milk production; 129 observations) and indigenous (160 observations) biotypes. Because of differences ( P 0.05). The equation from the development subset for dairy and indigenous goats was compared with the equation from the meat goat data set; there was a difference ( P 0.36 and 0.23, respectively), but the linear effect was significant ( P <0.01 and 0.05, respectively) when the cubic effect was removed from the model. For indigenous goats, quadratic and cubic effects were significant ( P <0.01), but the linear effect was not ( P =0.74); however, removal of the cubic effect resulted in a significant linear effect ( P <0.01). Therefore, simple linear regressions were performed. For each regression analysis, the residual (difference between actual and predicted values) for each treatment mean observation was compared with various multiples of the residual S.D. (rS.D.). Observations with residual greater than selected rS.D. were removed, and changes in R 2 and root mean square error (RMSE) were viewed. The rS.D. used to exclude observations was chosen on the basis of a moderate to appreciable increase in explained variability, with retention of the maximum number of observations. Observations removed were examined for possible reasons for high residuals ( Chatterjee et al., 2000 ). Final equations were tested for differences among biotypes in intercepts and slopes. Intercepts did not differ ( P =0.37) among biotypes; the slope for dairy goats was similar ( P =0.64) to that for indigenous goats, but the slope for meat goats differed ( P <0.01) from those for dairy and indigenous goats. Hence, data sets for dairy and indigenous goats were pooled, and the data for meat goats were analyzed separately. The pooled data set for dairy and indigenous goats (i.e., non-meat goats), including 275 treatment means and representing 2673 goats, was split into development and evaluation subsets by report. Data in the two subsets were made as homogeneous as possible for most important variables (e.g., MPI, ADG and mean BW) by exchange of observations from a small number of reports. Mean, minimum and maximum values for important variables are summarized in Table 4 . With the development subset, linear, quadratic and cubic effects of ADG on MPI were checked to justify the use of simple linear regression as described previously. The modified equation from the development subset was used to predict MPI in the evaluation subset. Observed MPI was regressed against predictions to determine whether the intercept and slope differed from 0 to 1, respectively. A comparison of prediction equations for meat and non-meat goats revealed similar intercepts ( P =0.25) and a difference in slopes ( P <0.01); therefore, a dummy variable, D ( D =0 for non-meat goats, D =1 for meat goats) was used to address the slope difference. The GLM model included ADG and the interaction of D and ADG. The final equation consisted of a common intercept for the three biotypes, a common slope for dairy and indigenous goats and a slope correction or adjustment term for meat goats. The intercept of the equation was considered the MP m and the slope MP g . 3 Results 3.1 Initial regressions 3.1.1 Meat goats The equation for the regression of MPI (g/kg 0.75 ) against ADG (g/kg 0.75 ) was (1) MPI =2.82( S.E. =0.401)+(0.428( S.E. =0.0310)× ADG ) (n=60;R 2 =0.77; RMSE =1.118) Although the R 2 of Eq. (1) was fairly high, removal of two observations with residuals greater than 2.0×rS.D. yielded a slightly greater R 2 : (2) MPI =2.55( S.E. =0.360)+(0.441( S.E. =0.0276)× ADG ) (n=58;R 2 =0.82; RMSE =0.989) Regression lines for Eqs. (1) and (2) are presented in Fig. 1 . The two excluded treatment mean observations were from the same study ( Soto-Navarro et al., 2004 ) with Boer×Spanish wethers. Although reasons for high residuals are not apparent, these observations entailed use of diets containing corn gluten and fish meals that are relatively high in ruminally undegraded protein and ADG was low compared with others in the same study. Based on Eq. (2) , preliminary estimates of MP m and MP g were 2.55 g/kg BW 0.75 and 0.441 g/g ADG, respectively. 3.1.2 Dairy goats The equation for the regression of MPI (g/kg 0.75 ) against ADG (g/kg 0.75 ) was (3) MPI =3.12( S.E. =0.397)+(0.295( S.E. =0.0276)× ADG ) (n=129;R 2 =0.47; RMSE =2.068) To improve fit of the model, six observations with residuals greater than 2.0 rS.D. were removed, and the resultant equation was (4) MPI =2.83( S.E. =0.344)+(0.299( S.E. =0.0238)× ADG ) (n=123;R 2 =0.57; RMSE =1.778) Regression lines for Eqs. (3) and (4) are presented in Fig. 2 . The six excluded observations were above the regression line with relatively high MPI (approximately 154 g/day) and intermediate ADG (approximately 161 g/day). There were no apparent unique characteristics of these observations with respect to other variables, such as mean BW, dietary CP concentrations of CP or forage, ME intake, genotype, gender, etc. Preliminary estimates of MP m and MP g for growing dairy goats from Eq. (4) were 2.83 g/kg BW 0.75 and 0.299 g/g ADG, respectively. 3.1.3 Indigenous goats The equation for the regression of MPI (g/kg 0.75 ) against ADG (g/kg 0.75 ) was (5) MPI =3.19( S.E. =0.253)+(0.306( S.E. =0.0349)× ADG ) (n=160;R 2 =0.33; RMSE =1.672) To improve the fit of model, eight observations with residuals greater than 2.0 rS.D. were removed, yielding the following...