Xiaorong Yang

Ph.D.

Qilu Hospital of Shandong University

Associate researcher

PLCO (Learn more about this study)

PLCO-890

Jan 13, 2022

The influence of biological factors and environmental exposure on the occurrence and prognosis of various cancers and the construction of prediction models

Cancer is a kind of chronic non-communicable disease that seriously threatens human health. While it brings huge suffering to patients, it also brings a huge disease burden to human beings. It is estimated that in 2020 alone, there were 19.3 million new cancers worldwide and about 10 million deaths caused by cancer. The occurrence and development, treatment response and prognosis of cancer are often related to a variety of internal biological factors and external environmental exposure factors. Revealing these key factors closely related to the occurrence and development of cancer and establishing an effective prediction model will not only effectively promote the development of cancer prevention, but also provide a basis for the selection of treatment methods and prognosis prediction. The primary goal of this proposal is to explore the influence of internal biological factors(age, biomarkers, genomics, etc.), external environmental factors (lifestyle, diet, surrounding environment, etc.) and change trajectories on the occurrence, development, treatment response and prognosis of various cancers based on all available information of the PLCO project. The second goal of our proposal is to build a model to predict the occurrence and development of cancer, treatment response, and prognosis. Therefore, we want to obtain all the data of the PLCO project, including all the questionnaire information, screening information, laboratory testing information, Genome-wide association (GWAS) data, etc. After gaining access to PLCO data, we plan to use Mendelian randomization, propensity score, mixed models and other methods to control potential confounding factors, use restricted cubic spline regression interpolation to analyze non-linear relationships, and use Cox proportional hazard regression model, logistic regression model, Poisson regression, joint model, age-period-cohort model and other methods to evaluate the relative risk of potential key factors, GWAS analysis, cluster analysis, LASSO regression, principal component analysis, Leave-one-out approach and other methods assess the relationship between high-throughput internal biological factors and cancer risk, treatment response and prognosis. Multiplicative interactions are evaluated by likelihood ratio tests of nested models with and without interaction terms. Apply nomogram, regression classification tree, adaptive boosting, bagging regression, random forest, support vector machine, nearest neighbor algorithm and machine learning methods in the random training data set to build prediction and screening models, and perform verification in the internal remaining data set.