Influence of Allele Frequency on Predicting Animal Phenotype Using Back-Propagation Artificial Neural Networks  

作　　者：LI Xuebin YU Xiaoling GUO Yunrui XIANG Zhifeng ZHAO Kun REN Fei 

机构地区：[1]College of Animal Science, Henan Institute of Science andTechnology, Xinxiang 453003, Henan, China

出　　处：《Wuhan University Journal of Natural Sciences》2011年第2期101-105,共5页武汉大学学报（自然科学英文版）

基　　金：Supported by the Scientific Research Starting Foundation for Doctors, Henan Institute of Science and Technology of China

摘　　要：To overcome the obstacle of the fascinating relation in predicting animal phenotype value, we have developed a neural network model to detect the complex non-linear relationships between the genotypes and phenotypes and the possible interactions that cannot be expressed with equations. In this paper, back-propagation neural network is used to discuss the influences of different allele frequencies on estimating the polygenic phenotype value. To ensure the precision of prediction, normalization was needed to train the prediction model. The results show that back-propagation artificial neural networks can be used to predict the phenotype value and perform very well in allele frequency from 0.2 to 0.8, when the allele frequency is very small （less than 0.2） or big （more than 0.8）; however, the prediction model was not reliable and the predicted value should be carefully tested.To overcome the obstacle of the fascinating relation in predicting animal phenotype value, we have developed a neural network model to detect the complex non-linear relationships between the genotypes and phenotypes and the possible interactions that cannot be expressed with equations. In this paper, back-propagation neural network is used to discuss the influences of different allele frequencies on estimating the polygenic phenotype value. To ensure the precision of prediction, normalization was needed to train the prediction model. The results show that back-propagation artificial neural networks can be used to predict the phenotype value and perform very well in allele frequency from 0.2 to 0.8, when the allele frequency is very small （less than 0.2） or big （more than 0.8）; however, the prediction model was not reliable and the predicted value should be carefully tested.

关 键 词：artificial neural network single-nucleotide polymorphism （SNP） HapMap project genomic breeding value molecular marker allele frequency 

分 类 号：Q347[生物学—遗传学] TP183[自动化与计算机技术—控制理论与控制工程]