Risk prediction models incorporating genetic, environmental and lifestyle factors for common and rare malignancies

Cancer has led to server disease burden worldwide. According to the data of the GLOBOCAN 2020, although the top 7 cancer types (breast, lung, colorectum, prostate, stomach, liver, cervix uteri) account for the 52% of all the newly diagnosed, the disease burden of other types of cancer still cannot be neglected. However, previous studies on risk factors of cancer mainly focused these common cancer types, and for rare cancer, such evidence is still lacking. Previous risk prediction models incorporated established genetic, environmental and lifestyle risk factors to estimate individual risk, and numerous models have been developed for cancer. However, most of previous models only focused on one specific type of cancer, and the large majority of them focused on the common cancer types. Epidemiology and etiology studies have demonstrated that some common factors may be associated with elevated risk of various cancer types, but with different magnitude of effect. Therefore, it is possible to develop a pan-cancer risk prediction model using the up-to-date machine learning technique incorporating a comprehensive collection of genetic, environmental and lifestyle risk factors, to estimate the individual risk for various types of cancer. The database of UK Biobank serves as a good platform for conducting such study. This project aims to investigate: (1) identification of disease-specific or shared genetic, environmental and lifestyle risk factors for common and rare malignancies; (2) development of effective disease-specific and pan-cancer risk prediction models using up-to-date machine learning techniques; (3) design suitable prevention strategies based on the individual risk profiles derived from the risk prediction models. The project is estimated to last for 36 months. The findings of this project will provide important reference for policymaking regarding the personalized and precision cancer prevention.