The following set of scatterplots show the -log₁₀ fold changes in p-value at individual SNPs when comparing GWAS of our default summary statistic (mean) to GWAS based on an alternative statistic (median, maximum or first available).

Please refer to the Methods section for a complete description. The x-axis corresponds the fold changes for the SNP in MGI and the y-axis corresponds to the fold changes for BioVU. Positive log-fold changes indicate that the alternative statistic yielded a smaller (more significant) p-value than using the mean as a summary statistic. The upper-right (green) quadrant plots SNPs that decreased in p-value in both cohorts for the alternative statistic. The lower-left (blue) quadrant plots SNPs that increased in p-value in both cohorts. The two remaining quadrants indicate SNPs with discordant changes in p-value between the cohorts. GWAS catalog SNPs are plotted in red, novel SNPs for a given lab (if applicable) are plotted in purple, and the remaining SNPs are LD-pruned (for plotting convenience) and plotted in black. The white diamond displays an empirical null distribution of fold changes for non-associated SNPs. The first 22 pages display the three alternative summary statistics (maximum value, median value, and first available measurement) for a single lab. The following six pages contain the analogous plots showing log fold change in p-values for the comorbidity model, which includes binary covariates for various comorbid diseases with the potential to impact lab measures, to a default analysis that does not account for comorbidities.

(PDF)