化学计量学模型预测中国泌乳奶牛瘤胃挥发性脂肪酸组成的精度分析

Accuracy Analysis of Prediction of Ruminal Volatile Fatty Acid Profiles in Chinese Lactating Dairy Cows by Stoichiometry Models

本研究旨在评估化学计量学模型预测中国泌乳奶牛瘤胃挥发性脂肪酸(VFA)组成的精度,分析影响模型预测精度的原因。本研究选择了3个经典的奶牛瘤胃VFA模型,分别为M UR模型、DIJ模型和BAN模型。试验数据来自我国主要科研团队的18篇论文,包括14篇SCI、3篇中文核心期刊和1篇未见刊的英文文章,论文数据含动物饲粮、体重、干物质采食量、饲料添加剂、VFA各组分比例。采用预测误差均方(MSPE)和一致性相关系数(CCC)2种分析方法对3个经典的奶牛瘤胃VFA模型的估算精度进行评估分析。结果表明:BAN模型乙酸比例估算精度最高(决定系数=0.140,P=0.007),预测误差均方根为6.8%,误差主要来自整体偏差的偏离(47.8%)。3个模型无法预测丙酸、丁酸、其他酸比例。总之,BAN模型预测乙酸比例的精度是3个模型中最高的,但预测精度仍然偏低,误差来源于整体偏差的偏离,迫切需要利用更多数据建立适合我国国情的VFA化学计量学预测模型。

This study was conducted to evaluate the accuracy of models to predict ruminal volatile fatty acids（ VFA） profiles in Chinese lactating dairy cows,and to analyze factors that affect the accuracy of the models.Three classical models of ruminal VFA stoichiometry were selected,which were MUR model,DIJ model and BAN model. The VFA data was selected from 18 articles of Chinese scientists,including 14 SCI articles,3 articles from Chinese Core Journals and 1 unpublished manuscript,and data included diet,body weight,dry matter intake,feed additives,VFA proportions. Mean squared prediction error（ MSPE） and consistent correlation coefficient（ CCC） methods were employed to evaluate the prediction accuracy of MUR model,DIJ model and BAN model. The results showed as follow： BAN model had the highest prediction accuracy of acetate proportion（ R2= 0.140; P = 0. 007,RMSPE = 6. 8%）,with overall bias being 47. 8%. Propionate,butyrate and other acids proportions could not be predicted by the above three models. In conclusion,BAN model has highest prediction accuracy to predict acetate molar proportion among the three models,but the prediction accuracy is still lowwith the error mainly coming from the overall bias,and it is necessary to make use of more data to establish a VFA stoichiometry prediction model suitable for Chinese national conditions.