FAQ

Q. When should I include covariates files of reference panel and what should be included as covariates?

A. XPASS follows the MQS approach (Zhou, X. (2017)) and uses the LD information from the entire chromosomes to estimate heritability and coheritability. This approach yields smaller standard error, but requires the population structures in the reference panel to be properly corrected by including the covariates (e.g., principal components). Otherwise, the estimated heritability and coheritability can be biased by the population structures (see here for an example). Therefore, we strongly suggest users to include covariates (e.g., principal components) when using XPASS although they are optional in the software.

Q. Can I use genotypes from 1000 Genomes project as reference panel for constructing PRS?

A. For demonstration, we use the easily accessible 1000 Genomes project genotypes as reference panels in the manual. The reference genotypes from 1000 Genomes project provided in our repository only contain 1.3 millions of SNPs from 377 EAS samples and 417 EUR samples. When used to estimate the heritability and coheritability, these reference panels are good enough. However, for PRS construction, the result of XPASS is slightly less accurate than what we reported in the AJHG paper. The performance of XPASS can be better when a GWAS dataset with a larger sample size (e.g.,n>2000) and more SNPs (e.g., 3M) is used as the reference panel. We strongly suggest that users use their own reference panels with sufficiently large sample sizes if available.