Extended Data Fig. 5: Association of CSF to plasma protein ratios with CSF TFRC and PDGFRB.

a,b, Volcano plots of linear regression associations between CSF to plasma ratios and CSF levels of PDGFRB or TFRC. For each protein, we regressed Ratio ~ CSF PDGFRB + CSF TFRC + age + sex. PDGFRB has a strong CSF pQTL (5:150128912:C:T)⁴¹; only participants homozygous for the majority C allele had CSF PDGFRB levels consistently above the limit of detection and were included in this analysis. Regression was performed in the Stanford (n = 64) and Knight (n = 73) cohorts separately and meta-analysis was performed using fixed-effect models. X axis: Estimated meta-analysis coefficient β for the indicated CSF protein. Y axis: Benjamini-Hochberg corrected q value. c,d, Bar plot showing the percentage of protein ratios positively and negatively associated with PDGFRB or TFRC (q < 0.05), by source. e, Venn diagram showing overlap among ratios with significant positive associations with age, PDGFRB, or TFRC. f, g) Scatterplot of age (x-axis) and CSF levels of PDGFRB and TFRC (y-axis). The estimated regression line depicting the association between age and protein levels is plotted, with 95% confidence intervals. P-values were derived from a fixed-effects meta-analysis combining age coefficient estimates from linear regression in Stanford and Knight cohorts, with sex as a covariate. h,i, Enrichment for UniProt domains within the set of peripheral proteins whose ratios are associated with PDGRFB (h) or TFRC (i). Two-sided P-values were calculated using Fisher’s exact test and Benjamini-Hochberg correction was applied.