Figure 3: Asymmetric functional relationships and negative genetic interactions.

(a) Empirical support for predicted functional asymmetry in gene essentiality⁴⁹ and, for non-essential genes, in the growth defect of single-gene knockout^20,21. For predicted asymmetric pairs (A→B), the fraction f_AB (f₀₁=n₀₁/(n₀₁+n₁₀)), where 0=non-essential or no fitness defect and 1=essential or substantial growth defect is expected to be larger than 0.5 (fraction in pairs with no asymmetry). (b) One example of predicted functional asymmetry with experimental support. ESCRT-I is composed of four members, Vps23p, Vps28p, Vps37p and Mvb12p. Mvb12p (blue subunit) is predicted to functionally dependent on Vps23p, Vps28p and Vps37p (red) and not vice versa. Structural studies revealed that the functionally dependent subunit, Mvb12p, is a structural stabilizer, which changes the ESCRT-I core complex (Vps23p, Vps28p and Vps37p) from a fan-shaped structure (upper panel) to an elongated structure (lower panel)⁵⁷. First, the dependency of Mvb12 on the core complex is supported by the fact that Mvb12p is unstable in cells lacking any of the other ESCRT-I subunits⁵⁸. Second, that Vps23p, Vps29p and Vps37p do not strictly depend on Mvb12p is validated by the findings that certain MVB sorting pathways, such as carboxypeptidase S (CPS) and Ste2 sorting, are effective even if Mvb12p is absent^58,59. However, loss of function of Vps23p, Vps28p or Vps37p results in a complete block of the MVB pathway⁵⁸. (c) Enrichment of negative genetic interactions in fan-in motif A–C pairs. The fraction of negative genetic interaction is defined as f=N_neg/(N_neg+N_pos+N_{noInteraction}). P-values in (a) and (c) were calculated with a one-tailed Fisher’s exact test and visualized with an asterisk (*) that stands for a P-value <0.05.