X-Wing (Cross-population Weighting) is a cross-population polygenic prediction framework only using GWAS summary statistics. It contains three methods:

• LOGODetect(LOcal Genetic cOrrelation DETECTor) (Also see here)
• PANTHER(Polygenic prediction using Bayesian ANnoTation-dependent HorsEshoe Regression)
• LEOPARD(LinEar cOmbination of PRS using gwAs summaRy statistics Data)

The X-Wing software can perform four tasks:

1. Improve genetic risk prediction in ancestrally diverse populations using X-Wing.
2. Identify small genome segments harboring cross-population local genetic correlation signals using LOGODetect.
3. Incorporate annotation into single and cross-population polygenic prediction using PANTHER.
4. Estimate the linear combination weights for multiple PRS by GWAS summary statistics alone using LEOPARD.

The layout of the scripts is

Please see the wiki for the detailed manual of X-Wing.

• May 26, 2022: Initial release.

Part of the code is adapted from PRS-CSx and XPASS. We thank Dr. Tian Ge and Dr. Mingxuan Cai for sharing their code.

The X-Wing posterior SNP effect size estimates in our manusciript are available at here.

If you use X-Wing, please cite

Miao, J., Guo, H., Song, G. et al. Quantifying portable genetic effects and improving cross-ancestry genetic prediction with GWAS summary statistics. Nat Commun 14, 832 (2023). https://doi.org/10.1038/s41467-023-36544-7

For questions and comments, please open a Github issue (preferred) or contact Jiacheng Miao at [email protected] or Hanmin Guo at [email protected].