Fig. 4: Association between protein abundance and radiographic OA severity after stratifying for biological sex in non-IPS regressed Combined data using logistic regression modelling.

Protein abundance was measured in 1322 samples (Combined: 1096 spun, 226 unspun), adjusted for spin-status (using ComBat) and then age. Volcano plot showing log odds ratios (logOR) per standard deviation change in protein expression for proteins associated with advanced radiographic status (KL grades 3-4) in the Combined dataset, with Benjamini-Hochberg adjusted p-values (padj), in A males (N = 623) and B females (N = 699). Proteins in red are positively associated, and those in blue negatively associated with advanced radiographic status (padj ≤ 0.05). Top 20 associated proteins in each direction, by padj, are labelled. In orange are proteins that replicated (significant at padj ≤0.05 and with effects in the same direction) in males and females, whereas white labelled proteins were only associated in the sex-specific set. C Scatter plot of logOR from logistic regression models of the associations between protein abundance and advanced radiographic disease status in males and females is shown with significantly associated proteins (at padj ≤ 0.05) in different groups shown in different colours (see key). Pearson correlation coefficient and p-value (unadjusted) are presented for the correlation between logOR generated in male-only and female-only analyses (Combined dataset). D Bubble plot of significantly enriched pathways (padj <0.05) using the Hallmark Gene set for proteins associated with advanced radiographic disease status by biological sex with and without additional adjustment for log (haemoglobin A protein expression). IPS intracellular protein score, SD standard deviation, logOR log odds ratio, padj adjusted p-value, SD standard deviation, NES normalised enrichment score. The full list of proteins is available in the Source Data file 7.