Computational Pipeline

For internal rearrangements of the complex (stoichiometry), we performed a separate analysis applying the R-package LIMMA (Linear Models for Microarray data analysis, [Ritchie et al., 2015]) using the complex-normalized protein abundances as input. Analogous to differential expression analysis, proteins showing a difference in their complex-normalized abundance relative to the other complex members were considered differentially expressed or stoichiometrically different between two given conditions. Contrasts were set accordingly to identify differences between male/female mice and high-fat/chow mice, respectively. For each complex, protein complex members were subjected to stoichiometric analysis; log2 fold-changes as well as p-values (moderated t-test) were collected. P-values were adjusted using the Benjamini-Hochberg procedure across all complexes and proteins. In case of q-value < 0.01 the corresponding protein was considered to be stoichiometrically changing in a given complex. The underlying statistical test is denoted as ‘LIMMA-based t-test’ throughout the main text. The corresponding fold-changes are highlighted in volcano plots in Figure 5B, and Figure S3C. The analysis was also performed for Reactome pathways, and can be readily applied to any specified protein set/module. To assess which complexes are affected in their stoichiometry as a whole, q-values of their individual components were combined using Fisher’s method. Lastly, the resulting combined p-values from all complexes were adjusted using the Benjamini-Hochberg method.