Q&A 11 How do you interpret GWAS model results with PCA covariates?

11.1 Explanation

Once a GWAS model is fitted using a phenotype (e.g., Plant height), a SNP, and population structure covariates (e.g., PC1–PC3), we interpret the results using the model summary. The key values to look for are:

• Estimate: The effect size of each variable
  
• Pr(>|t|): The p-value, used to determine significance
  
• R-squared: The proportion of variation in the trait explained by the model
  
• Residuals: The spread of errors not explained by the model

This example tests the association between SNP id1000007 and Plant height, adjusting for PC1 to PC3.

11.2 R Model Output Summary

Coefficient	Estimate	Std. Error	t value	Pr(>	t
(Intercept)	115.83448	1.04707	110.63	< 2e-16	***
id1000007	2.17767	1.54100	1.413	0.158
PC1	0.20502	0.02761	7.426	7.49e-13	***
PC2	-0.19534	0.04436	-4.404	1.39e-05	***
PC3	-0.29738	0.07399	-4.019	7.05e-05	***

Model Fit:

• Residual standard error: 18.63
• Degrees of freedom: 378
• R-squared: 0.2277
• Adjusted R-squared: 0.2195
• F-statistic: 27.86 on 4 and 378 DF
• Overall p-value: < 2.2e-16

✅ Takeaway: This SNP is not significant (p = 0.158), but PCs show strong association with plant height. Controlling for population structure is essential to avoid false signals in GWAS.