(A) Schematic representation of sequential cotranslational assembly in homomers. (B) Diagrammatic representation of the hypothesis test of the adaptive model of cotranslational assembly. (C) Area differences between the first and last translated interfaces in multi-interface heteromeric subunits across the species. Panels are ordered by the area cutoffs, 400, 1000, and 2000 Å2, which are satisfied if either the first or the last interface is larger than the given cutoff. Error bars represent standard error of the mean (SEM) and labels on bars show the number of proteins in each group. The p values were calculated with Wilcoxon signed-rank tests. (D) Bootstrap distributions of the area difference between the first and the last translated interfaces within two different categories. The first category (top two boxplots) is protein age, whereby yeast and human multi-interface heteromeric subunits are classified into ‘ancient’ and ‘more recent’ groups. In the second category (bottom boxplot), bacterial subunits are grouped based on whether or not they are encoded in operons. Positive values represent a larger first translated interface. The p values were calculated from 104 bootstrap resamples with correction for finite testing. (E) Example of an operon-encoded complex, the RecBCD nuclease (pdb: 5ld2). In the linear sequence of RecD, the interface with RecC is translated first, and that with RecB is last. The RecD:RecC interface is twice the area of the RecD:RecB interface, likely to promote cotranslational subunit recruitment. (F) Correlation between the relative distance of translational start points and the relative area difference of the first and last translated interfaces. Shaded lines represent the 95% confidence interval of the regression line. The Pearson’s correlation coefficient r, its p value, and the regression coefficient β1 are shown in the panels.