A robust and accurate method to detect MET Exon 14–skipping mutation in pan-cancer samples using an NGS-based RNAseq approach.

person Yongjie Wang The Affiliated Hospital of Qingdao University, Qingdao, China info_outline Yongjie Wang, Ren HengDa, Yiqun Zhang, Zhihua Pei, Dongliang Wang, Qiming Zhou, Jing Wang

Authors person Yongjie Wang The Affiliated Hospital of Qingdao University, Qingdao, China info_outline Yongjie Wang, Ren HengDa, Yiqun Zhang, Zhihua Pei, Dongliang Wang, Qiming Zhou, Jing Wang Organizations The Affiliated Hospital of Qingdao University, Qingdao, China, ChosenMed Technology (Beijing) Co.Ltd, Beijing, China, ChosenMed Technology (Beijing) Co., Ltd., Beijing, China, ChosenMed Technology (Zhejiang) Co., Ltd., Beijing, China Abstract Disclosures Research Funding Pharmaceutical/Biotech Company Beijing ChosenMed Laboratory Co., Ltd. Background: Met exon 14 skipping mutation(METex14) occurs in multiple cancer species, mainly in non-small cell lung cancer (NSCLC, 3~4%) and pulmonary sarcomatoid carcinoma (PSC, 22%). Several drugs expressed curative effect in previously treated patients who harboring METex14, such as savolitinib approved by NMPA, capmatinib and tepotinib licenced by FDA. Developing a quick and precise method which detect METex14 and help patients benefit from clinical drug treatment is a desiderative issue currently. Previous researches suggest that transcriptome-based RNA-seq method was superior to DNA-based NGS (exome, panel) and PCR method. Here We introduce a fast and accurate method to detect METex14 alternation in RNAseq data. Methods: We sequenced NGS-based Ribo-zero RNAseq (pair-end, 150bp×2) and METex14 ddPCR on 99 pan-cancer samples, of which 38 lung carcinoma (38%), 8 renal carcinoma (8%), 7 pancreatic carcinoma (7%). In 38 lung carcinoma, 35 NSCLC samples were consist by 29 lung adenocarcinoma (LUAD, 76%), 4 lung squamous carcinoma (LUSC, 10%) and 2 PSC (5%). Begin of the study, data quality control applied by Fastp and aligned to human hg19 genome by STAR subsequently. Next, MET exon14 related candidate soft-clipping reads were extracted from BAM files and translated to fastq file. Then we remapped the translated fastq file to hg19 genome with Blat using parameters ‘stepSize = 5, -repMatch = 2253, -minScore = 20, -minIdentity = 90’. Next, supporting reads were sequentially retained by: single end length longer than 18bp; block mapped reads of MET gene; the pair end reads mapped to genome if one end fall in the last 3bp of MET intron 13 and another end fall in at least initial 3bp of MET exon 15. Pair or single end support junction read, distance of junction site to reads end and site coverage were enrolled in a bayesian model to calculate METex14 confidence score (BMMETs). Threshold of our model was defined by the cut-point of max youden-index and 16 samples were validated METex14 by ddPCR. Finally, we would predict a sample with METex14 when score > = 0.752. Results: Our method reported 15 TP samples, 1 FP sample and 1 FN samples (AUC 0.97, sensitivity 0.88, specificity 0.98) while STAR-fusion detected 14 TR samples and 6 FP sample and 2 FN samples (AUC 0.715, sensitivity 0.64, specificity 0.73). BMMETs FP sample scored approximately 0.739 that close to the threshold. Conclusions: Strategy of screen reliable METex14 supporting reads and using bayesian model to calculate a METex14 confidence score could robustly and efficiently predict METex14 event in NGS-based RNAseq data.