Fig. 1

Study design and workflow. Study design and overall workflow of WGS analysis of germline DSVs and immune gene expression for cancer risk prediction and survival stratification. In total, 192 cancer patients (i)—comprised of 120 with colorectal cancer, 29 with endometrial cancer, 35 with ovarian cancer, and eight with breast cancer—were enrolled in the study group, and 499 non-cancer subjects (i) were included in the reference group. Genomic data, including WGS, gene expression, clinical outcome, and FCH, were collected. First, we used the PopDel method (ii) to detect DSVs and perform data preprocessing (ii) from the WGS analysis of all subjects. The cancer risk predictive model (iii) was built with an attention-weighted model. We also studied DSVs in familial cancer (iv). Second, we examined the relationship between DSVs and the tumor microenvironment (v). Immune gene expression data were normalized. We constructed an immune gene expression-associated DSV correlation matrix with the point-biserial correlation. Third, a machine learning method with a survival support vector machine (survival-SVM) and Kaplan–Meier survival analysis was applied to examine prognosis and survival (vi)