[ 摘要 ]

GBV-C/HGV在世界許多國家都有被發現的報告，台灣在一般族群有這方面的研究，但是僅有一篇文獻報告有關GBV-C/HGV在東部原住民�堣妓惘瘝v，卻無任何有關的分子流行病學研究和資料，故本論文探討GBV-C/HGV在東部原住民的盛行率以及流行的基因型與親源關係分析。

本論文以RT/nested PCR擴增5’UTR以及E2此二片段，並以三種不同的演算法(neighbor joining、maximum parsimony、maximum likelihood)進行親緣關係分析。在140個檢體中，5’UTR以及E2分別可檢測出4個及6個陽性反應的檢體，盛行率為2.8％(4/140)及4.2％(6/140)，與亞洲其他國家比較並沒有統計上顯著差異存在。親緣關係分析中5’UTR以及E2此二片段的陽性反應檢體均屬於第三型(Asian type)，且neighbor joining 及maximum likelihood方法其拓樸學均獲得統計上的支持。在病毒RNA盛行率分析發現，E2片段比5’UTR區域的敏感度高，而5’UTR區域比E2片段適合做親緣關係樹分析。

此外，我們亦實際運算證實，經過適當編輯的多重序列排序結果，會使得親緣關係分析更加正確，其統計值均有明顯上升。Bootstrapping 次數設定的多寡(100次及1000次)，亦會對統計值造成影響。1000次的bootstrapping 次數可提供更可信之統計值。

[ 英文摘要 ]

GB virus C (GBV-C)/hepatitis G virus (HGV) has been detected in eastern Taiwanese indigenes. However, only the prevalence of GBV-C/HGV viral RNA was reported, but not genotyping and their phylogenetic relationship. We therefore investigated the prevalence, genotypes, and phylogeny of GBV-C/HGV in eastern Taiwanese indigenes to complete their molecular epidemiological data. We used RT/nested PCR to amplify the 5’ untranslated region and the second envelope gene (E2) and followed by direct sequencing. Among 140 samples, 4 were found to be GBV-C/HGV RNA positive (2.8％) for 5’UTR and 6 (4.2％) for the E2 region respectively. These sequences were analyzed using different phylogenetic methods (neighbor joining、maximum parsimony、maximum likelihood), since the algorithms for constructing phylogenetic trees of different methods are based on different evolutionary assumptions; a consistent topology with all methods increases its reliability. All these isolates belonged to the genotype 3 (Asian type) cluster, in phylogenetic analysis, base on 5’UTR region and E2 segment. The tree topology was supported statistically significant by neighbor joining and maximum likelihood methods. The detection of E2 gene is more sensitive than 5’UTR in the prevalence study; however, 5’UTR is more suitable to study in phylogenetic analysis. Furthermore, we also found a well-edited sequence alignment would greatly improve the results of confident in tree topology; and a significant effect on different times of bootstrap re-sampling datasets (100 and 1000 times) was observed.