A Model Study of In Silico Proficiency Testing for Clinical Next-Generation Sequencing
A Model Study of In Silico Proficiency Testing for Clinical Next-Generation Sequencing
复制标题
DOI：
10.5858/arpa.2016-0194-cp
复制
发表时间：
2016-10-01
影响因子：
4.6
通讯作者：
Pfeifer, John D.
中科院分区：
文献类型：
--
作者：
Duncavage, Eric J.;Abel, Haley J.;Pfeifer, John D.
背景。-目前，下一代测序分析的大多数能力测试挑战是基于方法的能力测试调查，这些调查使用来自表征参考样本的DNA来测试测试的湿台和生物信息学/干台方面。基于方法的能力验证调查受到天然存在或可以引入单个DNA样品的突变数量和类型的限制。为了解决这些局限性，探索一种计算机模拟能力测试模型，其中对来自单个充分表征的标本的序列数据进行电子操作。使用Illumina TruSeq和Life Technologies AmpliSeq面板富集来自美国病理学家学院参考基因组的DNA，并分别在MiSeq和Ion Torrent平台上测序。将所得数据进行计算机诱变处理，并以10%至50%的变异等位基因分数（VAF）添加26种变异，包括单核苷酸变异、缺失和二核苷酸取代。参与的临床实验室下载了这些文件，并使用其临床生物信息学管道进行分析。使用AmpliSeq/Ion Torrent和/或TruSeq/MiSeq的实验室参与了2项调查。平均而言，实验室在26种变异中识别出24.6种（95%），在22种变异中识别出21.4种（97%），VAF大于15%。未报告假阳性呼叫。最常遗漏的变异是VAF小于15%的单核苷酸变异。在这两种挑战中，报告的VAF一致性非常好，模拟VAF和平均报告VAF之间的中位绝对差异小于1%。结果表明，计算机能力验证是一种可行的方法为基础的能力验证方法，并证明，目前的下一代测序生物信息学跨临床实验室的灵敏度和特异性高。
Context.-Most current proficiency testing challenges for next-generation sequencing assays are methods-based proficiency testing surveys that use DNA from characterized reference samples to test both the wet-bench and bioinformatics/dry-bench aspects of the tests. Methods-based proficiency testing surveys are limited by the number and types of mutations that either are naturally present or can be introduced into a single DNA sample.Objective.-To address these limitations by exploring a model of in silico proficiency testing in which sequence data from a single well-characterized specimen are manipulated electronically.Design.-DNA from the College of American Pathologists reference genome was enriched using the Illumina TruSeq and Life Technologies AmpliSeq panels and sequenced on the MiSeq and Ion Torrent platforms, respectively. The resulting data were mutagenized in silico and 26 variants, including single-nucleotide variants, deletions, and dinucleotide substitutions, were added at variant allele fractions (VAFs) from 10% to 50%. Participating clinical laboratories downloaded these files and analyzed them using their clinical bioinformatics pipelines.Results.-Laboratories using the AmpliSeq/Ion Torrent and/or the TruSeq/MiSeq participated in the 2 surveys. On average, laboratories identified 24.6 of 26 variants (95%) overall and 21.4 of 22 variants (97%) with VAFs greater than 15%. No false-positive calls were reported. The most frequently missed variants were single-nucleotide variants with VAFs less than 15%. Across both challenges, reported VAF concordance was excellent, with less than 1% median absolute difference between the simulated VAF and mean reported VAF.Conclusions.-The results indicate that in silico proficiency testing is a feasible approach for methods-based proficiency testing, and demonstrate that the sensitivity and specificity of current next-generation sequencing bioinformatics across clinical laboratories are high.