講演情報
[O25-4]マルチオミクス機械学習による食道癌の化学療法効果予測因子
○笹川 翔太1, 加藤 寛章3, 長岡 孝治2, Todd Johnson1, 前嶋 和紘1, 大川 裕貴1, 垣見 和宏2, 安田 卓司3, 中川 英刀1 (1.理化学研究所 生命医科学研究センター がんゲノム研究チーム, 2.東京大学 医科学研究所, 3.近畿大学 医学部 外科)
Esophageal squamous cell carcinoma (ESCC) is one of the most aggressive cancer and is primarily treated with platinum-based neoadjuvant chemotherapy (NAC). However, biomarkers to predict NAC sensitivity and their response mechanism in ESCC remain unclear. We performed RNAseq and shallow WGS on 121 pre-treatment ESCC biopsy specimens, comprehensively profiled immuno-signatures and copy-number signatures, and created a model to predict responsiveness toNAC using machine learning (Random forests). In our model, Chromosome 12q,neutrophils, CNSignature6, and smoking were found to be negative predictorsfor non-responders to NAC. We validated an important role of neutrophils inESCC response to NAC in the mice experiment. We also demonstrated that specific copy number alterations and copy number signatures in the ESCC genome were significantly associated with NAC response and machine learning algorismon the multi-omics data predicted NAC response with AUC=0.8 in another validation cases (n=20). Machine learning technique on WGS and RNAseq data is useful to mine biomarkers and its diagnostic prediction model created by thesedifferences may be a new indicator of therapeutic capability.