Table 1.
Hypothetical pathways and predicted values of experimental parameters.
Figure 1.
Theoretical model used to interpret epistasis in signal-responsive pathways.
Lines ending in circles indicate influences that can be activating or repressing. Dotted lines indicate Boolean relationships defined in Eq. (4). Variables and influences are defined in the text (αI is omitted for clarity).
Table 2.
Predicted values of experimental parameters used for inferring pathway architecture and gene order.
Table 3.
Definitions of influences for hypothetical pathways based on the values of the trait T and the difference trait D.
Figure 2.
Canonical yeast galactose utilization pathway (adapted from[26]).
Gray arrows indicate cellular processes likely impacted by GAL gene deletions. Abbreviations: intracellular galactose (galIN), galactose-1-phosphate (gal1P), glucose-1-phosphate (glu1P), uridine diphosphate (UDP), UDP-glucose (UDPglu), UDP-galactose (UDPgal).
Figure 3.
(A) Relative fitness for single GAL gene deletions, in presence (red) and absence (blue) of galactose. Asterisk marks significant effect (t-test p-value <0.05). (B) Signal-dependent genetic interactions (green: δI>0, red: δI<0, t-test p-value <0.05). (C) Inferred causal relationships. (D) Transitive reduction of the network in (C).
Figure 4.
Quantitative pathway inference using fitness traits.
Relationships inferred between the signal (gal) and GAL genes are shown with values of their signal-specific influences (95% confidence) on fitness (T). Pathways involving GAL6 are omitted.
Figure 5.
Analysis of reporter expression traits.
(A) Relative traits for single GAL gene deletions, in presence (red) and absence (blue) of galactose. Asterisk marks significant effect (t-test p-value <0.05). (B) Signal-dependent genetic interactions (green: δI>0, red: δI<0, t-test p-value <0.05). (C) Inferred causal relationships. (D) Transitive reduction of the network in (C).
Figure 6.
Quantitative pathway inference using expression traits.
Relationships inferred between the signal (gal) and GAL genes are shown with values of their signal-specific influences (95% confidence) on expression (T).
Figure 7.
Analyses of true (TPR) and false positive rates (FPR) for inference of causal relationships among GAL genes.
(A, B) TPR (circles) and FPR (diamonds) for analysis of fitness and expression, respectively. (C) TPR (circles) and FPR (diamonds) for combined analyses. (D) Receiver operating characteristic curves for fitness, expression and combined analyses.