基于高通量测序技术的拷贝数变异筛选分析流程的建立及应用

Application of copy number variation screening analysis process based on high-throughput sequencing technology

目的基于高通量测序的CNVs分析流程(PICNIC)与常规拷贝数变异微阵列比较基因组杂交检测方法的检测效率和结果分析。方法纳入2016年1月1日至2017年12月31日复旦大学附属儿科医院分子诊断中心基于临床需要的、取得患儿家长知情同意的、同时送检了微阵列比较基因组杂交检测和高通量测序分析的病例。以微阵列比较基因组杂交为金标准,PICNIC为待测标准,微阵列比较基因组杂交检测采用安捷伦人类基因组CGH微阵列180K试剂盒,并经其软件进行数据处理和拷贝数变异检测。选择重复片段>500kb,缺失片段>200 kb的CNVs数据分析,经过筛选后,结合患儿表型人工进行结构评判,致病/可能致病(P/LP)为检测到的CNVs已知的表型与患儿表型相符合;临床意义未明(VUS)为检测到的CNVs已知的表型与患儿表型不完全相符合。PICNIC分析流程,从同一测序批次的BAM文件开始,经过外显子覆盖深度计算、质控筛选、CANOES计算CNVs评分并提供候选CNVs。后续从基因水平和区域水平二方面对CNVs进行注释和筛选。比较2种方法间检出率及其敏感性。结果 113例同时送检了微阵列比较基因组杂交检测和PICNIC测序分析流程,平均年龄2岁,发育迟缓82例,惊厥16例,孤独症5例,先天性心脏病3例,遗传咨询病例7例。微阵列比较基因组杂交检测到P/LP为76例,VUS为16例; PICNIC检测到P/LP为92例,VUS为21例;微阵列比较基因组杂交检测到的P/LP和VUS病例,均包含在PICNIC检测到的病例中,16例微阵列比较基因组杂交检测VUS结论的CNVs被PICNIC纳入P/LP,敏感度100%(95%CI:94%～100%),特异度100%(95%CI:81%～100%),阳性预测值82.6%(95%CI:73%～89),阴性预测值56.8%(95%CI:40%～72%)。微阵列比较基因组杂交检测到的446个CNVs中,PICNIC也检测到190个,在PICNIC到的236个CNVs中,微阵列比较基因组杂交检测也检测到190个。结论 PICNIC分析流程对于P/LP的CNVs,具有100%的特异性和敏感性,可用于CNVs的临床检测。该方法的建立及临床推广对于进一步挖掘高通量捕获测序数据具有重要意义。

Objective To compare the detection results of the copy number variations analysis pipeline based on high-throughput sequencing(PICNIC)and conventional array comparative genomic hybridization.Methods We enrolled patients underwent both aCGH detection and high-throughput sequencing in Children's Hospital of Fudan University from January 1,2016 to December 31,2017.Informed consent was obtained from the parents.We have established an operational pipeline(PICNIC)which detect,annotate and prioritize CNVs from raw high-throughput sequencing data combined with clinical information for disease diagnosis.The CNVs identification between aCGH platform and the PICNIC pipeline were compared to prove the utility value of PICNIC.The aCGH platform used the Agilent's SurePrint 180K Kit for experiment and specific softwares for data processing and copy number variation identification.Detected CNVs with duplications>500 kb and deletions>200 kb were remained for further analysis.With the manually evaluation of the patient's clinical phenotypes and the CNV's function,Pathogenic/Likely-pathogenic(P/LP)were concluded for the CNVs that reported phenotypes were in accordance with the patients phenotypes while Variants of Unknown Significance(VUS)for the not exactly matched ones.The PICNIC analysis pipeline started with the BAM file generated from the sequencing data.After the exon coverage depth calculation and the quality control,CANOES was used for the CNVs detection.The detected CNVs were further annotated and filtered for both gene level and regional level.The detection rate and sensitivity of these two methods were compared.In this study,we performed aCGH and PICNIC on 113 samples and compared the result of CNVs detection between these two methods.Results Altogether,113 cases underwent both aCGH detection and PICNIC analysis were enrolled,with an average age of 2 years old.The patients including 82 developmental delay cases,16 seizures cases,5 autism cases,3 congenital heart disease cases and 7 genetic counseling cases.The aCGH test detected 76 P/LP CNVs and 16 VUS CNVs while PICNIC detected 92 P/LP CNVs and 21 VUS CNVs.All of the P/LP and VUS CNVs detected by aCGH were consistently detected by PICNIC and 16 VUS CNVs detected by aCGH were concluded as P/LP by PICNIC.Taking aCGH as the gold standard,the detection of PICNIC showed that sensitivity was 100%(95%CI:94%-100%),specificity was 100%(95%CI:81%-100%),positive predictive value was 82.6%(95%CI:73%-89%),and negative predictive value was 56.8%(95%CI:40%-72%).Among the 446 aCGH detected CNVs,PICNIC verified 190 of them;among the 236 CNVs detected by PICNIC,aCGH identified 190 of them.Conclusion The CNVs analysis pipeline based on high throughput sequencing(PICNIC)demonstrated 100%specific and sensitive to Pathogenic/Likely pathogenic CNVs,and could be used in clinical detection.The establishment of this method and its clinical popularization were of great significance for further high-throughput capture sequencing data mining and re-analyzing.