On October 27, the reporter learned from the Tumor Hospital of Tianjin Medical University that Professor Hao Jihui, the president of the hospital, has innovatively developed a non-invasive detection model based on circulating free DNA (cfDNA), which provides a new scheme for early screening of pancreatic cancer by integrating multi omics characteristics and machine learning. The relevant research results were published in the international journal Cancer Discovery. Pancreatic cancer is the most malignant cancer in clinic, known as the "king of cancer". The 5-year survival rate of patients is only 11%. Due to the lack of effective early screening methods, most patients are diagnosed in the late stage and miss the best treatment opportunity. The existing diagnostic methods such as magnetic resonance imaging and endoscopic ultrasound are difficult to apply to large-scale screening due to their high cost and invasiveness; However, the commonly used tumor marker CA19-9 has limited specificity and sensitivity in early diagnosis, which cannot meet clinical needs. Hao Jihui's team successfully developed an early screening model for pancreatic cancer by integrating the copy number variation, fragment length distribution, fragment chain bias and other multiple omics characteristics of cfDNA, combined with machine learning technology. In the training queue containing 467 samples and the validation queue containing 352 samples, the model demonstrated excellent screening performance, significantly better than the screening ability of traditional tumor marker CA19-9. More importantly, the team also verified the clinical value of the model in a prospective cohort of 1926 people with diabetes and obesity. Six cases of pancreatic cancer were successfully detected by this model at early screening, with a detection rate of 75%. All cases were detected at early stage (stage 0, stage I or stage II), while only one case was detected by using tumor marker CA19-9. Compared with imaging examinations, this model can detect lesions 45-298 days in advance (median 227.5 days), which has won valuable time for patients to undergo early intervention treatment. According to the simulation analysis of the team, if the model is applied to clinical screening, the diagnosis rate of pancreatic cancer patients in Phase I can be increased from 6.22% to 85.58%. Under the ideal condition that all patients are screened at an early stage for effective treatment, the 5-year survival rate can be increased from 7.59% to 33.34%, and the absolute survival rate due to early diagnosis can be improved to 25.75%. The study further carried out Monte Carlo simulation on 100000 hypothetical population. Compared with the use of tumor marker CA19-9, this model can detect 20 additional pancreatic cancer patients in 100000 hypothetical population, with a sensitivity increase of 8.22%, and the positive predictive value and accuracy rate are significantly better. Wang Xiuchao, the main member of the project and deputy chief physician of Tianjin Medical University Cancer Hospital, said that this innovative method is expected to promote the transformation of pancreatic cancer from "late diagnosis" to "early intervention", and bring an important breakthrough in the field of pancreatic cancer diagnosis and treatment. (New Society)