️ Episode 184: High-Accuracy Multiethnic XGBoost for Skin Cancer Identification

In this episode of PaperCast Base by Base, we explore a large-scale study that builds a risk factor–based XGBoost model using the All of Us cohort to accurately identify patients with skin cancer across diverse ancestries.

Study Highlights:
Analyzing more than 400,000 participants, the authors quantify independent associations between genetic ancestry, lifestyle, social determinants of health, prior cancer history, and use of PDE5A inhibitors with skin cancer risk. They compare traditional logistic regression against gradient-boosted trees and show that logistic models have low precision for case identification, motivating a non-linear approach. The resulting multiethnic XGBoost model achieves high accuracy for identifying patients with any skin cancer, with F1 scores of 0.903 in individuals of European ancestry and 0.810 in non-European groups. SHAP importance and interaction analyses reveal strong non-linear effects of age and genotype principal components, and suggest that genetic and socioeconomic factors contribute more heavily to predictions in younger individuals.

Conclusion:
A multiethnic, non-linear model that integrates genetics, lifestyle, social determinants, and medication exposures can substantially improve early identification of skin cancer patients across ancestries, offering a precision-medicine tool to help reduce outcome disparities.

Reference:
D’Antonio M, Gonzalez Rivera WG, Greenes RA, Gymrek M, Frazer KA. A highly accurate risk factor–based XGBoost multiethnic model for identifying patients with skin cancer. Nature Communications. 2025;16:9542. https://doi.org/10.1038/s41467-025-64556-y

License:
This episode is based on an open-access article published under the Creative Commons Attribution 4.0 International License (CC BY 4.0) – https://creativecommons.org/licenses/by/4.0/

Support:
If you'd like to support Base by Base, you can make a one-time or monthly donation here: https://basebybase.castos.com/