Please select your Region.
Please select your Region.
Aug 28, 2024
A Study for the Development of a Pancreatic Cancer Diagnostic Method Using miRNA in Blood
Comprehensive Analysis of miRNA in Blood can Discriminate the Development of Pancreatic Cancer with High Accuracy
A group led by Dr. Munenori Kawai, Associate Professor Akihisa Fukuda, and Professor Hiroshi Seno of the Department of Gastroenterology and Hepatology at the Kyoto University Graduate School of Medicine has revealed that comprehensive analysis of miRNA in blood can discriminate the development of pancreatic cancer with high accuracy, in collaboration with ARKRAY, Inc. at 14 facilities including Kindai University and Kyoto Prefectural University of Medicine under the Moonshot Research and Development Program. Pancreatic cancer is often asymptomatic in the early stages and is difficult to detect, making it representative of cancers with a poor prognosis. In this study, the group constructed a discrimination model to discriminate pancreatic cancer with automated machine learning using data obtained from comprehensive analysis of miRNAs in the blood of pancreatic cancer patients and non-cancerous healthy control subjects using a next-generation sequencer. The model showed that pancreatic cancer patients could be discriminated with high accuracy from non-cancerous healthy controls in an independent validation cohort. Patients with early-stage asymptomatic pancreatic cancer could also be discriminated from non-cancerous healthy controls, indicating that this technology may lead to the early detection of pancreatic cancer.
The results will be published online in the British Journal of Cancer on August 28, 2024, local time.
The number of deaths from pancreatic cancer is on the rise as the fourth leading cause of cancer-related deaths nationwide. It is known as one of the cancer types with the poorest prognoses, with a five-year survival rate of 8.5%. Due to various efforts in diagnosis and treatment, prognoses for pancreatic cancer patients are gradually improving, but remain poor. About half of pancreatic cancers are diagnosed at stage 4 with distant metastasis at the time of detection, and the five-year survival rate is significantly reduced to less than 5% if distant metastasis is present. On the other hand, while pancreatic cancer that is localized to the pancreas and potentially resectable within a margin of 1 cm at the time of diagnosis accounts for only a few percent of all pancreatic cancers, a ten-year survival rate of more than 90% has been reported in such cases. This means that early detection and diagnosis is important to improve the prognosis of pancreatic cancer. However, many of the early-stage pancreatic cancers have few specific symptoms and are asymptomatic, and there are no useful biomarkers to identify early-stage pancreatic cancer. For these reasons, early detection and diagnosis of pancreatic cancer are currently very difficult. Therefore, there is a need for new, less invasive tools to discriminate and diagnose pancreatic cancer at an early stage.
MicroRNAs (miRNAs) are short RNAs composed of 18 to 24 bases and are known to regulate gene expression through extracellular secretion by extracellular vesicles such as exosomes and through uptake into other cells. Previous studies have suggested that miRNAs in blood could be a new biomarker for pancreatic cancer, but there has never been a validated study focusing on early-stage--and in particular asymptomatic--pancreatic cancer patients.
A group led by Dr. Munenori Kawai, Associate Professor Akihisa Fukuda, and Professor Hiroshi Seno of the Department of Gastroenterology and Hepatology at the Kyoto University Graduate School of Medicine has comprehensively measured miRNAs in the blood of 212 pancreatic cancer patients and 213 non-cancerous healthy control subjects using a next-generation sequencer to construct a pancreatic cancer discrimination model which can discriminate pancreatic cancer patients from non-cancerous healthy controls with automated machine learning in collaboration with ARKRAY, Inc. at 14 facilities, including Kindai University and Kyoto Prefectural University of Medicine, and its discrimination performance was validated by an independent validation cohort.
Automated Machine Learning (AutoML) technology was used to create the pancreatic cancer discrimination model, and the expression data of 100 miRNAs were used to construct an ensemble model combining eight algorithms.
When comparing the performance of the discrimination model between pancreatic cancer patients and non-cancerous healthy controls using miRNAs with that of CA19-9 (an existing tumor marker), the results showed that the area under the curve (AUC) of CA19-9 was 0.88, while the AUC of the miRNA model was 0.94, and the AUC of the miRNA + CA19-9 model was 0.99, indicating that pancreatic cancer patients can be discriminated from non-cancerous healthy controls with high accuracy (Fig. 1 and Fig. 2). Furthermore, when limited to stage 0 and stage 1 early-stage pancreatic cancer, the AUC was 0.81 for CA19-9, while the AUC was 0.92 for the miRNA model and 0.98 for the miRNA + CA19-9 model, indicating that even patients with early-stage pancreatic cancer can be discriminated from non-cancerous healthy controls with high accuracy (Fig. 3). Furthermore, in asymptomatic early-stage pancreatic cancer, the sensitivity of CA19-9 was 0.29, while the sensitivity of the miRNA model was 0.48, and the sensitivity of the miRNA + CA19-9 model was 0.67, indicating that this technology could lead to early detection and diagnosis of pancreatic cancer.
(Click to enlarge)
The pancreatic cancer discrimination model in this study was constructed using miRNA expression data measured with the NextSeq 550, a next-generation sequencer manufactured by Illumina. However, the validation cohort measured using the Ion GeneStudio S5, a next-generation sequencer manufactured by Thermo Fisher Scientific Inc., was also able to discriminate pancreatic cancer, indicating that it is a highly robust model for detecting differences in miRNA assays.
This study comprehensively measured miRNAs in blood, created pancreatic cancer discrimination models using automated machine learning, and validated their discrimination performance using an independent validation cohort. As a result, it was shown that pancreatic cancer patients can be discriminated from non-cancerous healthy controls with high accuracy. Patients with early-stage pancreatic cancer can also be discriminated from non-cancerous healthy controls, indicating that this study may be a new, less invasive screening method that can detect pancreatic cancer at an early stage. In the future, research and development will be continued in collaboration with ARKRAY, Inc. to promote societal adoption of the technology used in this study, such as verifying whether the miRNA-based pancreatic cancer discrimination model constructed in this study is useful for high-risk populations with risk factors for pancreatic cancer.
This work was funded by joint research funds of ARKRAY, Inc. and supported in part by Grants-in-Aid for JSPS Scientific Research (19H03639, 19K16712, 19K22619, 20H03659, 21K19480, 23K21432, 23K27582, 23H02891, 24K02438), AMED Project for Cancer Research and Therapeutic Evolution (P-CREATE: 18cm0106142h0001, 19cm0106142h0002, 20cm0106177h0001, 21cm0106177h0002, 21cm0106283h0001, 22cm0106283h0002), AMED Project for Promotion of Cancer Research and Therapeutic Evolution (P-PROMOTE: 23ama221326h0001, 24ama221326h0002, 24ama221515h0003), AMED Bridging the fundamental mechanism of aging and the effective treatment of age-related disease associated with impaired functional system (AMED-PRIME: 21gm6010022h0004), Moonshot Research and Development Program (JPMJMS2022-1, JP22zf0127009), JST Program on open innovation platform for industry-academia co-creation (COI-NEXT: JPMJPF2018), JST Fusion Oriented Research for disruptive Science and Technology (FOREST: 23719768), Takeda Science Foundation, Princess Takamatsu Cancer Research Fund, Astellas Foundation for Research on Metabolic Disorders, Daiichi Sankyo Foundation of Life Science, The Yasuda Medical Foundation, The Uehara Memorial Foundation, the Naito Foundation, and Kyoto University.
1. Next-Generation Sequencer (NGS)
This device allows simultaneous reading of base sequences in small DNA fragments ranging from tens of millions to tens of billions, and has been used in many pieces of scientific research as well as in cancer gene panel testing for cancer genomic medicine in recent years. It also enables comprehensive measurements of small molecule RNA such as miRNA with high reproducibility.
2. AutoML (Automated Machine Learning)
This technology automates machine learning processes, enabling automated optimization of data preprocessing and hyperparameters of machine learning algorithms, as well as algorithm searches with high prediction accuracy. It has been receiving increased attention for its efficiency in model building.
3. AUC (Area Under the Curve)
This is the area under the ROC (Receiver Operating Characteristic) curve, which is used to evaluate the performance of binary classification models. The closer the AUC value is to 1, the higher the discrimination performance. A value of 0.5 indicates that discrimination is completely random.
Pancreatic cancer is difficult to detect or diagnose at an early stage, and detection rate at stage 0 or 1 is less than 10%. Pancreatic cancer is expected to have a long-term prognosis if it can be operated on at stage 0 or stage 1, but there are currently no existing biomarkers that can detect and diagnose pancreatic cancer at an early stage. During this joint study with ARKRAY, Inc. at 14 facilities, the group developed a pancreatic cancer discrimination model using automated machine learning by comprehensively measuring miRNAs in blood. As this group validated the model using an independent validation cohort, the results indicated that pancreatic cancer patients could be discriminated from non-cancerous healthy controls with a high degree of accuracy. Patients with early-stage pancreatic cancer can also be discriminated from non-cancerous healthy controls, indicating that this study may be a new, less invasive screening method that can detect pancreatic cancer at an early stage. This study is a data analysis of miRNAs in blood collected from the largest number of early-stage pancreatic cancer patients ever. In particular, it is important in that it is the first to show that even asymptomatic early-stage pancreatic cancer patients can be discriminated from non-cancerous healthy controls. In the future, we plan to pursue further research and development in collaboration with ARKRAY, Inc. toward the societal adoption of the technology gained through this study.
Title: Early detection of pancreatic cancer by comprehensive serum miRNA sequencing with automated machine learning
Authors: Munenori Kawai, Akihisa Fukuda*, Ryo Otomo, Shunsuke Obata, Kosuke Minaga, Masanori Asada, Atsushi Umemura, Yoshito Uenoyama, Nobuhiro Hieda, Toshihiro Morita, Ryuki Minami, Saiko Marui, Yuki Yamauchi, Yoshitaka Nakai, Yutaka Takada, Kozo Ikuta, Takuto Yoshioka, Kenta Mizukoshi, Kosuke Iwane, Go Yamakawa, Mio Namikawa, Makoto Sono, Munemasa Nagao, Takahisa Maruno, Yuki Nakanishi, Mitsuharu Hirai, Naoki Kanda, Seiji Shio, Toshinao Itani, Shigehiko Fujii, Toshiyuki Kimura, Kazuyoshi Matsumura, Masaya Ohana, Shujiro Yazumi, Chiharu Kawanami, Yukitaka Yamashita, Hiroyuki Marusawa, Tomohiro Watanabe, Yoshito Ito, Masatoshi Kudo, Hiroshi Seno
*Corresponding author
Published in: British Journal of Cancer
DOI：10.1038/s41416-024-02794-5
Kyoto University Hospital, Kindai University Hospital, University Hospital Kyoto Prefectural University of Medicine, Osaka Red Cross Hospital, Japanese Red Cross Wakayama Medical Center, Japanese Red Cross Otsu Hospital, Medical Research Institute KITANO HOSPITAL , Tenri Hospital, Shiga General Hospital, Hyogo Prefectural Amagasaki General Medical Center, Kyoto Katsura Hospital, Kobe City Nishi-Kobe Medical Center, Shinko Hospital, Red Cross Hospital Takatsuki
