Fig. 4: Adjustment for PILRA-associated PAV improves inter-platform correlation of measures and strengthens association with age.
From: Cross-ancestry comparison of aptamer and antibody protein measures
A Olink vs. SomaScan standardized (“std”) and inverse normal transformed normalized (“invn”) protein expression (NPX) and relative fluorescence units (RFU), respectively, without adjustment for PAV rs1859788 (chr7:100374211:A:G; p.Arg78Gly) (left) and with adjustment for the PAV (right). Shaded bands indicate the 95% confidence interval for the linear regression fit. B SomaScan (pink) and Olink (blue) PILRA measures by genotype in 1889 unrelated MESA participants with protein measures and whole genome sequencing data available. Data are presented as the median protein levels (center line) with the 25th and 75th percentiles as box bounds; whiskers extend to 1.5× the interquartile range, and points beyond represent outliers. C pQTL summary statistics (generated with linear regression under an additive model, two-sided test of significance) and effect allele frequencies per MESA ancestry-group. cis-pQTL were significant at traditional genome-wide significance threshold (p < 5 × 10−8). This variant is most common among AMR and EAS ancestry groups. D Inter-platform protein measure correlation per ancestry without and with PAV adjustment. E Forest plot of age association statistics generated from linear regression of age versus protein (two-sided test of significance), using SomaScan and Olink measures without and and with PAV adjustment (“PAV-adj”) in 1889 participants. Boxes represent estimated beta coefficients, and horizontal lines show 95% confidence intervals. Associations were significant at \(\alpha\)= 0.05 correcting for 2708 probe pairs tested (p < 1.8 × 10−5). Source data is available in Supplementary Data 4, 5 (C), and 6 (D, E). Source data for (A, B) is individual-level data available through dbGap and MESA.
