基於文字探勘排序來自外顯子組資料之基因變異點

Abstract

次世代基因定序提供我們一個相當快速的方法去獲得DNA的定序資料，病人的DNA經過quality checking, alignment, 變異點偵測以及變異點分析之後，我們可以獲得外顯子 (Exome) 中單點核苷酸變異 (Single Nucleotide Variants, SNVs) 的資料，這些資料包含病人體細胞中所有的SNVs。然而雖然基因突變可能會導致人類發生疾病、影響對抗病原體的生理機制、改變對於化學物質或藥物的正常反應，但是僅僅只有少數的變異會具有致病性或是將導致生理機制的失調及疾病。為了更快速地找尋真正具有致病性且與臨床上症狀相關之變異點，我們使用自行開發的MViewer，NGS的輔助判讀程式，來做資料的前處理，以及基於文字探勘技術的變異點排序工具 (Variants Prioritizer) 排序所有可能的變異點。 MViewer整合來自多個資料庫中對於外顯子變異點的說明，同時根據具有致病性變異點的共同特徵，例如在某個特定族群中變異點發生的頻率較低等，或是其他條件來篩選出可能性較高變異點。接著我們會根據病人臨床上的症狀及病名給予其一些關鍵字，這些關鍵字與變異點候選將會做為本研究開發的變異點排序工具 (Variants Prioritizer) 之輸入，變異點排序工具會透過分析GeneReviews、Online Mendelian Inheritance in Man (OMIM) 與PubMed上的摘要文章來判斷關鍵字與變異點候選之間的關聯性，最後輸出依據關聯性高到低的變異點候選列表，藉此來找出最有可能導致疾病及與病人臨床病徵最相關的變異點。 經過分析32人及其父母的外顯子變異點資料後，我們成功地使用MViewer將原本平均多達30,000個變異點降至500個基因，變異點排序工具亦成功地找出目標變異點，目標變異點出現在候選列表前十名的成功率高達87.5%。這些導致疾病的基因包含CD3E, EIF2B5, COQ4, MOCS2, ISPD, NGLY1, COQ4, PEX1, PITX2, KCNQ5, RYR2, SCN8A, PEX7, C3, TRPC6, TNNI3, CPS1, EDN1, SLC22A5, PAH, RAG1, EYA1, IFT122, ATP7A, BRCA1, ELN, FRAS1, COL2A1, ALDH18A1, MAF, SUOX以及CTLA4。由上述可知MViewer以及變異點排序工具在外顯子定序資料的分析上是相當快速且有幫助的。

Next Generation Sequencing (NGS) provides us a rapid way to get the DNA sequence data. After quality checking, alignment, variants detection and annotation, we will get the exome data of single nucleotide variants (SNVs), which are variations in the single nucleotide without any limitation of frequency and may arise in somatic cells. Variants in the DNA sequences of humans can affect how humans develop diseases and response to pathogens, chemicals, drugs, and other agents, but only one or very few are expected to be pathogenic or significant for the relevant disorder. In order to narrow down further towards culprit disease genes, we use MViewer, which can analyze the SNVs data with various annotations from different genome database, as our preprocessing tools to decrease the numbers of SNVs. Then, in this research, we are planning to describe a novel tool, Variants Prioritizer, based on text-mining to use articles from GeneReviews, Online Mendelian Inheritance in Man (OMIM) and abstracts from PubMed to rank SNVs from MViewer and keywords from electronic medical records (EMRs) to get the most possible disease-causing candidates. After testing 32 trio sequencing data, we successfully decreased the number of candidate SNVs from average about 30,000 to 500 candidate genes with MViewer. Variants Prioritizer succeeds in 87.5% of the cases to locate the target gene in the top 10 ranking list. Disease causing genes identified included CD3E, EIF2B5, COQ4, MOCS2, ISPD, NGLY1, COQ4, PEX1, PITX2, KCNQ5, RYR2, SCN8A, PEX7, C3, TRPC6, TNNI3, CPS1, EDN1, SLC22A5, PAH, RAG1, EYA1, IFT122, ATP7A, BRCA1, ELN, FRAS1, COL2A1, ALDH18A1, MAF, SUOX and CTLA4. Therefore, MViewer and Variants Prioritizer are very powerful in WES reading. Currently Variants Prioritizer is used to help clinical exome reading.