猪饲料有效能值预测模型的构建  

Construction of Prediction Models of Feedstuffs Effective Energy Values for Swine

作　　者：潘晓花[1] 杨亮[1] 庞之洪[1] 王建芬 熊本海[1] 

机构地区：[1]中国农业科学院北京畜牧兽医研究所动物营养学国家重点实验室,北京100193 [2]延庆县动物卫生监督管理局,北京102100

出　　处：《动物营养学报》2015年第5期1450-1460,共11页CHINESE JOURNAL OF ANIMAL NUTRITION

基　　金：基本业务费课题(2013ywf-zd-3);动物科学与动物医学数据共享平台课题;国家科技支撑计划课题(2014BAD08B05)

摘　　要：为探索饲料常规成分及碳水化合物组分与饲料有效能值之间的关系方程,本研究以NRC第11版《猪营养需要量》中发布的122套饲料营养成分表为基础,将饲料中11种基础成分[6项常规成分：干物质、粗蛋白质（CP）、粗纤维（CF）、粗脂肪（EE）、酸性醚提取物、粗灰分（ash）;5项碳水化合物组分：淀粉（ST）、中性洗涤纤维（NDF）、酸性洗涤纤维（ADF）、半纤维素、酸性洗涤木质素]作为自变量,将饲料中的消化能（DE）、代谢能（ME）及净能（NE）作为因变量,采用SAS软件中的REG过程,分别建立不同性质饲料、自变量的不同组合与DE、ME及NE之间的回归关系方程,并以相关系数（R2）及变异系数（CV）作为评价回归模型的优劣。研究表明,有效能值与CP、ST及纤维类指标显著或极显著相关（P〈0.05或P〈0.01）。将所有饲料作为研究对象时,饲料的DE、ME及NE与上述11种基础成分之间建立的普适性回归模型预测效果较差。当将14种玉米及其加工产品形成子集时,建立饲料基础营养成分与DE、ME及NE的关系方程分别为7、6和7套（P〈0.05）,且3组回归模型R2分别为0.632 8~0.772 3、0.646 9~0.684 9和0.670 5~0.822 1,CV分别为6.61%~8.40%、6.58%~7.34%和6.21%~8.27%;当将13种大豆及其加工产品形成子集时,共建立饲料基础成分与DE、ME关系方程分别有3和4套,回归模型R2分别为0.907 1~0.926 9、0.890 7~0.922 3,CV分别为5.40%~6.09%、5.79%~6.78%,NE与基础营养成分指标之间无法建立具有营养学意义的有效回归方程。对于同类饲料中具有相同自变量组合的DE及ME预测模型而言,两者之间的差异主要是自变量CP的系数上,且CP部分对ME的正效应低于DE,这保证模型预测的ME低于DE。同时选用本研究构建的适宜模型,补充了NRC第11版成分表中第97（去皮大豆粕,低寡糖,浸提）、101（全脂大豆,高蛋白质）及102号（全脂大豆,低寡糖）饲料�This study was conducted to establish the relation equations between feedstuffs＇ chemical composi- tions, carbohydrate fractions and effective energy value. Based on intensively analysis of the indexes changed in the NRC 11th ed. swine feedstuff composition table, and the table was selected as data source to predict DE, ME and NE indirectly by basic chemical compositions and their different combinations, which were 6 kinds proximate nutrients [ dry metter （ DM）, crude protein （CP） , crude fiber （CF） , ether extract （EE） , acid hy- drolysis ether extract （ AEE）, ash ] and 5 kinds of carbohydrate （CHO） composition [ starch （ ST）, neutral detergent fiber （ NDF）, acid detergent fiber （ ADF）, hemicellulose （ HC ）, acid detergent lignin （ ADL ） were considered as independent variables, while digestible energy （DE）, metabolizable energy （ME） and net energy （NE） were treated as dependent variables, using REG process of SAS to set up these relationship equa- tions for different feedstuff groups and different independent variable combinations. The correlation coefficient （ R2 ） and coefficient of variation （CV） were used to evaluate the fitness of models. The results showed as fol- lows： when considering all feedstuff as a group, universally applicable prediction models couldn＇ t be estab- lished between DE, ME, NE and feedstuffs＇ chemical compositions. Further research found that when consid- ering corn and it＇ s by-products as a subset, 7, 6 and 7 models were built for DE, ME and NE, respectively, and their R2 were 0.632 8 to 0.772 3 （CV was 6.61% to 8.40%）, 0.646 9 to 0.684 9 （CV was 6.91% to 7.34%） and 0.670 5 to 0.822 1 （CV was 6.22% to 8.28%）. Three and four models were built for DE and ME respectively when considering soybean and it＇ s by-products as a subset, and their R2 were 0. 907 ] to 0.926 9 （CV was 5.40% to 6.09%）, 0.890 7 to 0.922 3 （CV was 5.79% to 6.78%）, no linear regression existed between NE and basic chemica

关 键 词：NRC饲料成分表 消化能 代谢能 净能 预测模型 

分 类 号：S828[农业科学—畜牧学] S816.17[农业科学—畜牧兽医]