Whether to include reference genome in PCA for association analysis?

Question

[Phoenix Mu]

When doing QC for GWAS, genotype from a reference panel like 1000 Genomes is typically used to calculate PCA together with our own data to find individuals with unmatched population ancestry.

My question is if I want to calculate PCs used for adjusting population admixture in my GWAS, do I also need to include a reference panel in my PCA?

Thanks

Answer 1

[zx8754]

Use reference panel to get PCs, this step is useful to filter out outlier samples. Then exclude outlier samples and reference samples, and re-run PCA. These are the PCs we want to use for analysis.

See this paper for more info:

• Data quality control in genetic case-control association studies

[AH]

If we have samples from one ancestry (south asian ) do we need to remove outliers based on 1000 genomes or we should remove outliers by PC within the sample. In other words if the samples are from single ancestry why do we need to to remove samples using 1000 genomes ? There might be admix within the SAS like Punjabis , Tamil or Pathans, but not EU, AFR. An alternative approach can be the we construct PCs using Southasian data from 1000 genomes .