AJS-4 AI consultation for NGS results

Abstract

In 2019, the two types of cancer gene profiling (CGP) test have been approved by PMDA in Japan. We have developed a novel CGP system by amplicon exome sequence targeting 160 cancer genes: “PleSSision” in 2017. We have examined more than 2000 solid cancer patients including the clinical examination, and the success rate of sequence was 98%, and detection rate of actionable gene mutation was more than 60 %. Moreover, we have recently launched “PleSSision-Exome” covering around 20000 genes. In this innovative system, we established high volume and deep clinical sequence for cancer tissue with reasonable cost as laboratory examination. For 18 months, we have examined around 150 cancer patients and the success rate of exome sequence was more than 90 %, however, around 15% of them could reach the genotype-matched treatment, although as much better result than the reimbursed CGP test. Current bioinformatics pipeline for NGS result is usually consist of multiple steps including read mapping, curation for SNVs, Ins/Del and CNVs, and annotation. During these processes, semi-automatic curation system already can exclude SNPs and irregular reads, and call the significant gene alterations, although the background genomic instability including whole genome duplication and/or low quality of specimen sometimes read the misinterpretation. The annotation for identified gene alteration is much chaotic part. Global cancer databases such as ClinVar and COSMIC will help us to interpret clinical significance for each gene alteration, however excessive and discrepant information sometimes confuse the oncologist to recommend the genotype matched treatment. AI technology is now expected to provoke the dramatic improvement for such clinical annotation, however still we have a lot of challenges to be concur. In this symposium, based on our experience during the operation of “PleSSision-System”, challenges of precision cancer medicine using AI technology will be discussed.