Figure 4. Estimating E_c for real proteomes.

(A) Computed E_c as a function of optimal growth temperature (OGT). The correlation coefficient is 0.60 ± 0.11. More stable proteins, lower E_c, are found in thermophiles. (B) Same plot as in A but with E_c computed using Eq.4. The correlation coefficient is now significantly smaller 0.39 ± 0.10. E_c is overestimated because the alphabet size is overestimated; moreover, the E_c estimate in A discards the components of interactions that are the same for all residues. (C) Composition difference between organism with lowest OGT (black) and highest OGT (white).