This repository contains R code used to perform the data analysis descrived in the manuscript:
Novel protein markers of androgen activity in humans: proteomic study of plasma from young chemically castrated men. https://elifesciences.org/articles/74638
Differentially expressed proteins were determined by doing one-way repeated measures ANOVA (R function: ezANOVA{ez}) to reveal overall differences between conditions (three time points) and differences between individual conditions were detected by performing a post-hoc test based on pairwise t-test (two-tails and paired) (R function: pairwise.t.test{stats}). The ‘pairwise.t.test’ function utilized the proteins with significant overall changes (ANOVA p-value < 0.05) to perform pairwise comparisons between conditions while corrected for multiple pairwise testing. Proteins with adjusted p-values (‘fdr’ method) < 0.05 following the pairwise t-test were considered significant
Receiver operating characteristic (ROC) analysis to select proteins capable to discriminate between normal and low Testosterone. (Healthy human model)
A stepwise regression (method: backward) to select the best combination of markers that predict the odds of being low Testosterone. Bootstrap resampling with replacement method was applied to assess consistency of predictors selected with the stepwise regression.
ROC analysis to discriminate patients with MetS, IR, CVRLP, DM or LBD within the cohort infertile men, using the expression level of the candidate biomarkers and Testosterone hormone. Based on the ROC curves, the DeLong's test (paired) was applied in R (roc.test {pROC} function) to compare the AUCs of the candidate biomarkers vs the AUC of Testosterone levels to discriminate the above mentioned pathological conditions.