Improvement in the Determination of HIV-1 Tropism Using the V3 Gene Sequence and a Combination of Bioinformatic Tools
Improvement in the Determination of HIV-1 Tropism Using the V3 Gene Sequence and a Combination of Bioinformatic Tools
复制标题
DOI：
10.1002/jmv.21425
复制
发表时间：
2009-05-01
影响因子：
12.7
通讯作者：
Garcia, Federico
中科院分区：
文献类型：
--
作者：
Chueca, Natalia;Garrido, Carolina;Garcia, Federico
使用基于V3序列的生物信息学工具评估HIV趋向性与表型趋向性测定提供的结果相关性较差，特别是在识别X4病毒时。这可能是使用CCR5拮抗剂的一个障碍。结合几种生物信息学工具的算法可能会改善与表型倾向结果的相关性。共有200个来自HIV-1亚型B的V3序列，在几个已知表型结果的数据库中可用，用于评估七种不同的生物信息学工具（PSSM, SVM C4.5决策树生成器和C4.5， PART， Charge Rule和Geno2pheno）的敏感性和特异性。确定了最好的预测性生物信息学工具，并建立了一个结合其中几个工具的模型。在200个参考序列中，SVM和geno2-pheno检测X4病毒的灵敏度最高，分别为98.8%和93.7%；但特异性较低，分别为62.5%和86.6%。对于R5病毒，PSSM和C4.5给出了相同的结果，并且优于其他生物信息学工具（95.7%的灵敏度，82%的特异性）。当这四种工具中的三种结果一致时，以表型倾向试验结果为参考，预测R5或X4病毒的敏感性和特异性均超过90% （AUC: 0.9701; 95% Cl: 0.9358-0.9889）。该算法结合了四种不同的生物信息学工具（SVM, geno2pheno， PSSM和C4.5），改进了HIV倾向的基因型预测，值得进一步评估，因为它可能在临床实践中被证明是有用的筛选策略。中华医学杂志，2009,31(1):763-767。(C) 2009 Wiley-Liss,Inc。
Assessment of HIV tropism using bioinformatic tools based on V3 sequences correlates poorly with results provided by phenotypic tropism assays, particularly for recognizing X4 viruses. This may represent an obstacle for the use of CCR5 antagonists. An algorithm combining several bioinformatic tools might improve the correlation with phenotypic tropism results. A total of 200 V3 sequences from HIV-1 subtype B, available in several databases with known phenotypictropism results,were used to evaluate the sensitivity and specificity of seven different bioinformatic tools (PSSM, SVM C4.5 decision tree generator and C4.5, PART, Charge Rule, and Geno2pheno). The best predictive bioinformatic tools were identified, and a model combining several of these was built. Using the 200 reference sequences, SVM and geno2-pheno showed the highest sensitivity for detecting X4 viruses (98.8% and 93.7%, respectively); however, their specificity was relatively low (62.5% and 86.6%, respectively). For R5 viruses, PSSM and C4.5 gave the same results and outperformed other bioinformatic tools (95.7% sensitivity, 82% specificity). When results from three out of these four tools were concordant, the sensitivity and specificity, taking as reference the results from phenotypic tropism assays, were over 90% in predicting either R5 or X4 viruses (AUC: 0.9701; 95% Cl: 0.9358-0.9889). An algorithm combining four distinct bioinformatic tools (SVM, geno2pheno, PSSM and C4.5), improves the genotypic prediction of HIV tropism, and merits further evaluation, as it might prove useful as a screening strategy in clinical practice. J. Med. Virol. 81:763-767,2009. (C) 2009 Wiley-Liss,Inc.