HEART: Heterogeneity-aware Empowerment via Adaptive Robust Transfer

HEART tests variant-trait associations in a (small) target population while borrowing strength from a (large) source population, without assuming that effects are shared.

Details

The workflow has two steps:

1. heart_assoc() computes, for every variant, the target-only estimator, an auxiliary estimator (pooled, doubly robust or source-only), and the decorrelated auxiliary estimator, yielding asymptotically independent p-values \((p_{tar,j}, p_{de,j})\). With only summary statistics, heart_meta() gives the meta-analytic counterpart.

2. heart_test() combines the p-value pairs through a four-group mixture model into the HEART statistic and estimates its composite null distribution, providing FDR and FWER control.

heart() runs both steps on individual-level data.

See also

Useful links:

• https://github.com/haoyu-yang001/HEART

• https://haoyu-yang001.github.io/HEART/

• Report bugs at https://github.com/haoyu-yang001/HEART/issues

Author

Maintainer: Haoyu Yang haoyuyang@hsph.harvard.edu

Authors:

• Ruoyu Wang

• Xihong Lin xlin@hsph.harvard.edu