Fig. 5: Nonlinear concentration-phenotype functions differentially transform dominance and epistasis.

a Linear, concave, convex and sigmoidal linking functions are used to transform protein concentrations into phenotypes. b Interaction scores based on the additive expectation for double mutants within- or between alleles with linear (Model 2), concave, convex, or sigmoidal protein concentration – phenotype relationships. c With vs. without nonlinear linking function comparisons of interaction scores based on the additive expectation. d Between- vs. within-allele double mutants’ interaction scores based on the additive expectation, with nonlinear linking functions. e Distribution of interaction scores based on additive expectation before and after nonlinear linking functions. The green arrow indicates the distribution shifting towards negative values while the magenta arrow indicates the distribution shifting towards positive values; the arrowheads point at the range after applying the nonlinear linking functions to the phenotype. f, g, h Dominance scores before and after nonlinear linking functions (f), distribution (g), and with vs. without nonlinear linking function comparisons (h).