Accurate classification and hemagglutinin amino acid signatures for influenza A virus host-origin association and subtyping
Accurate classification and hemagglutinin amino acid signatures for influenza A virus host-origin association and subtyping
复制标题
DOI：
10.1016/j.virol.2013.11.010
复制
发表时间：
2014-01-20
期刊：
影响因子：
3.7
通讯作者：
Sherif, Fayroz E.
中科院分区：
文献类型：
--
作者：
ElHefnawi, Mahmoud;Sherif, Fayroz E.
为了追踪甲型流感病毒的宿主来源，基于HA蛋白对甲型流感病毒的宿主来源分类和特征进行了研究。建立了甲型流感病毒各亚型及其主要宿主(人、禽、猪)的隐马尔可夫模型(HMM)、决策树和关联分类。寻找宿主和亚型特异的HA蛋白特征。识别了出现在不同亚型病毒中的宿主相关特征。对基于ROC曲线和对氨基酸类关联规则的支持度和置信度的分类模型进行了评估。特征选择后，基于HA亚型的主机分类准确率在91.2%~100%之间。鸟类-宿主来源的特定宿主类别关联规则给出了更好的支持度和置信度评级，其次是人类来源，最后是猪来源。这一发现表明猪宿主的特异性较低，可能表明它有能力混合不同的菌株。(C)2013 Elsevier Inc.保留所有权利。
Host-origin classification and signatures of influenza A viruses were investigated based on the HA protein for tracking of the HA host of origin. Hidden Markov models (HMMs), decision trees and associative classification for each influenza A virus subtype and its major hosts (human, avian, swine) were generated. Features of the HA protein signatures that were host-and subtype-specific were sought. Host-associated signatures that occurred in different subtypes of the virus were identified. Evaluation of the classification models based on ROC curves and support and confidence ratings for the amino acid class-association rules was performed. Host classification based on the HA subtype achieved accuracies between 91.2% and 100% using decision trees after feature selection. Host-specific class association rules for avian-host origins gave better support and confidence ratings, followed by human and finally swine origin. This finding indicated the lower specificity of the swine host, perhaps pointing to its ability to mix different strains. (C) 2013 Elsevier Inc. All rights reserved.