survkl: An R Package for Transfer-Learning–Based Integrated Cox Models

survkl provides flexible and efficient tools for integrating external risk scores into Cox proportional hazards models while accounting for population heterogeneity. The package enables robust estimation, improved predictive accuracy, and user-friendly workflows for modern survival analysis.

Key Features

• Integration of External Risk Scores
  Seamlessly incorporate predictions from any external risk model (e.g., polygenic hazard scores, deep-learning–based risk scores).

• Population Heterogeneity Adjustment
  Corrects for distributional differences between the external model’s training population and your study population.

• Efficient Computation
  Designed for high-dimensional and large-scale survival datasets.

• Improved Estimation and Prediction
  Demonstrated gains in estimation efficiency and predictive accuracy.

• Built-In Cross-Validation
  Automated selection of tuning and penalization parameters.

Installation

Note: This package is under active development. Please report any issues you encounter.
Requires R ≥ 4.0.0.

Install from CRAN:

install.packages("survkl")

Or install the development version from GitHub:

require("devtools")
require("remotes")
remotes::install_github("UM-KevinHe/survkl")

Detailed Tutorial

Full package documentation and parameter explanations: here

Getting Help

If you encounter problems or bugs, please contact us:

• ybshao@umich.edu
• lfluo@umich.edu
• xhliuu@umich.edu
• junyiqiu@umich.edu
• kevinhe@umich.edu

References

1. Wang, D., Ye, W., Zhu, J., Xu, G., Tang, W., Zawistowski, M., Fritsche, L. G., & He, K. (2023). Incorporating external risk information with the Cox model under population heterogeneity: Applications to trans-ancestry polygenic hazard scores. arXiv:2302.11123.

2. Luo, L., Taylor, J. M. G., Wang, D., & He, K. (2024). Flexible Deep Learning Techniques for Cox Models with Data Integration.