Leveraging clinical and genetic risk factors for risk prediction of eight cancers in the UK biobank.

BACKGROUND: Models with polygenic risk scores (PRSs) and clinical factors to predict risk for different cancers have been developed. However, these models have been limited by the PRS-derivation methods and the incomplete selection of clinical variables.

METHODS: We used UK Biobank (UKBB) to train the best PRSs for eight cancers (bladder, breast, colorectal, kidney, lung, ovarian, pancreas, and prostate) and select relevant clinical variables from 733 baseline traits through extreme gradient boosting (XGBoost). Combining PRSs and clinical variables, we developed Cox proportional hazards models for risk prediction in these cancers.

RESULTS: Our models achieved high prediction accuracy for eight cancers with AUCs ranging from 0.618 (95% CI 0.581-0.655) for ovarian cancer and 0.831 (95% CI 0.817-0.845) for lung cancer. Additionally, our models could identify individuals at a high risk of developing cancer. For example, the risk of breast cancer for subjects in the top 5% score quantile was nearly 13 times greater compared to subjects in the lowest 10%. Furthermore, we observed a higher proportion of high-PRS individuals in the early-onset group, but a higher proportion of high clinical-risk individuals in the late-onset group.

CONCLUSION: Our models demonstrated the potential to predict cancer risk and identify high-risk individuals with great generalizability to different cancers. Our findings suggested that the PRS model is more predictive for the cancer risk of early-onset patients than for late-onset patients, while the clinical risk model is more predictive for late-onset patients. Meanwhile, combing PRS and clinical risk factors have overall better predictive performance than using PRS or clinical risk factors alone.