INFERENCE OF EVOLUTIONARY CHANGES OF WITHIN-HOST VIRUS

我们已经提出了一种基于距离的序列链接法，用于纵向分析宿主内病毒的进化，并成功地重建了来自未接受药物治疗的患者的24个HIV-1 env基因序列的纵向系统发生树。然而，由于抗病毒药物的选择压力，正在接受药物治疗，特别是高效抗逆转录病毒治疗(HARRT)的患者的病毒进化要复杂得多。为了估计这个复杂的过程，我们将信息准则引入到我们的方法中，对原有的算法进行了很大的改进。我们使用基于互信息准则开发的C4.5(Quinlan，1993)软件包来获得识别病毒样本与所用药物组合之间关系的决策树。我们将修改后的算法应用于HAART下连续三年从单个患者身上连续采样的HIV-1蛋白酶基因的大数据集(189个序列)。得到的决策树表明，在整个观察期内，总有两个或两个以上的病毒亚群具有不同的频率和不同的特征突变。此外，通过估计决策树分类的每个病毒亚群的同义(Ds)和非同义(D_N)替换率，我们发现，当不同的药物使用时，施加在病毒上的选择压力随着时间的推移而变化很大，但病毒可能经常处于积极的达尔文选择下进化。上述结果已在包括国际会议(SMBE2004，ISMB2004)在内的多次会议上公布，并获得了第18届日本艾滋病研究学会年会的“优秀主题奖”

We have previously proposed an algorithm, distance-based sequential-linking method, for longitudinal analysis of within-host viral evolution and successfully reconstructed a longitudinal phylogenetic tree of 24 HIV-1 env gene sequences obtained from patients who had not received drug treatment. However, viral evolution in patients who are undergoing drug therapy, especially HARRT (Highly Active Anti-Retroviral Therapy), is much more complicated because of the selective pressure from the anti-viral drugs. To estimate this complicated process, we largely improved the previous algorithm by introducing information criterion into our method. We employed C4.5 (software package, Quinlan, 1993), which is developed based on mutual information criterion, to obtain the decision tree that identifies the relations between virus samples and drug combinations administered. We applied the revised algorithm to a large data set (189 sequences) of HIV-1 protease gene serially sampled from a single patient under HAART over three years. The obtained decision tree showed that there were always two or more viral subpopulations with different frequencies and also different characteristic mutations during the whole observation period. Moreover, by estimating the synonymous (ds) and nonsynonymous (d_N) substitution rates of each viral subpopulation classified by the decision tree, we found that the selective pressure exerted on the virus is highly variable over time when different drugs are administered, but the virus may often be under positive Darwinian selection to evolve.The above results have been presented at several meetings including international meetings (SMBE2004, ISMB2004), and received the "Excellent Theme Prize" from the 18th annual meeting of the Japanese society for AIDS research