Figure 5. Comparison of double mutant mutation outcome and genetic interactions in cis and trans.

(A) Cartoon illustrating how the cis library differs from the trans one. NNS, whole codon substitution. Asterisk, point mutation. (B) Scatter plot between Average PPI scores of identical pairs of mutations at the same positions in cis and trans. (C) Distribution of double mutants PPI scores in the original cis and trans libraries or after matching their single mutant effects distributions. Error bars represents 95% confidence intervals around the mean of 1000 sub-samplings when matching the two libraries. (D) Stacked bar-plot showing the proportion of the non-random variance in double mutant PPI scores that is not accounted for by the multiplicative model, explained by the thermodynamic model and the residual structural genetic interactions. Error bars represent the standard error of the mean. (E) Proportion of significant positive and negative genetic interactions in the two original libraries. See Figure 5—figure supplements 2, 3, 7 and 9 for sub-sampled libraries with matched single mutant effect distributions.

Figure 5—figure supplement 1. PPI scores for the cis library.

(A) Correlation between PPI scores from the three biological replicates of the cis mutation library selection. Contains both single (n = 190) and double (n = 17,290) mutations. Variants with less than 10 UMIs in any of the three replicate input samples or without any UMI in any of the replicate output samples were filtered out. (B) Correlation between single amino acid change PPI scores in the trans and cis libraries for all mutations reachable through a single nucleotide substitution.

Figure 5—figure supplement 2. Proportion of cis and trans double mutants classified as strengthening, intermediate effect or severely detrimental using different score thresholds.

The libraries where sub-sampled to match the distribution of the single mutant PPI scores as described in the method section. (A) Distribution of single mutants PPI scores before (solid lines) after (bars) sub-sampling. Mutants 1 and 2 correspond respectively to Fos and Jun in the trans library or to Fos single mutants randomly attributed to mutant 1 or 2 in each sub-sample of the cis library. For the original cis library, the distribution corresponds to pooling mutants 1 and 2 together dividing the counts by 2. (B) Double mutants stronger than WT. (C) Double mutants severely detrimental. (D–E) Double mutants of intermediate effect, varying the lower and upper thresholds, respectively.

Figure 5—figure supplement 3. Fitting the thermodynamic model on both the trans and cis data.

(A) Proportion of total variance explained in both datasets when fitting the model on either the trans or the cis data only. The models were fitted on the pooled replicates. Percentage of variance explained were then calculated for each replicate independently. (B) Distributions of the proportion of variance explained in 1000 sub-samplings to match the distribution of single mutant effects between the two libraries, showing that the model fitting does not depend on the differences in effect size distributions between the original libraries. (C) Average proportion of the non-random variance in double mutants PPI scores not accounted for by the multiplicative model and explained by the thermodynamic model in the 1000 sub-sampled libraries. The model was fitted on the two original libraries pooled together. The percentage of variance explained was then calculated for each replicate independently. 95% confidence intervals around the mean of the 1000 sub-samplings were not plotted because they are negligible (0.0028 for trans and 0.00058 for cis).

Figure 5—figure supplement 4. Significant structural genetic interactions in cis and trans.

(A) Correlation of structural genetic interaction scores between the three replicates of each library. Genetic interaction scores are from the thermodynamic model fitted on the two original libraries pooled together. (B) Volcano plot of the p-values of each double mutant as a function of the corresponding average genetic interaction scores. Data from the original libraries (not sub-sampled). Red lines represent the thresholds that were used to define the strong and significant genetic interactions used to compute the enrichment in structural features. Vertical lines, magnitude thresholds. Horizontal line, p-value threshold that corresponds to a FDR of 0.2.

Figure 5—figure supplement 5. Comparisons of the patterns of structural genetic interactions in cis and trans.

(A–B) Enrichment (Fisher’s exact test) for strong positive (green) or negative (purple) genetic interactions at different distance thresholds between the two mutated positions (A) and between different position types (B). These enrichments were also calculated after filtering out variants with 100 or less reads in any of the replicates input (bottom barplots on both panels), showing that the trends are the same even though the significance is lower with the stringent filter due to the smaller size of the dataset, especially for the trans library. The trans data analyzed here is restricted to amino acid substitutions reachable through single nucleotide changes. n = 17,290 and 14,118 for cis and trans, respectively, when filtering at 10 input reads, and 15,198 and 2,807, respectively, when filtering at 100 input reads. *, FDR < 0.1. **, FDR < 0.01. ***, FDR < 0.001. n.s., non significant.

Figure 5—figure supplement 6. Comparisons of the pairs of positions enriched in cis and trans.

(A) Pairs of positions significantly enriched in positive (green) or negative (purple) genetic interactions (Fisher’s exact test, FDR = 10%). Each pair of positions is also classified according to its heptad position type and the distance between the two positions (bottom matrix, yellow cells). (B) Heatmap (top) and distribution (bottom) of percentage of significantly positive (left) or negative (right) genetic interactions per pairs of position. The trans data analyzed here is restricted to amino acid substitutions reachable through single nucleotide changes.

Figure 5—figure supplement 7. Robustness of the enrichments in structural features when sub-sampling to match single mutant effects.

Enrichments calculated from one of the sub-samples picked randomly. (A–B) Enrichment (Fisher’s exact test) for strong positive (green) or negative (purple) genetic interactions at different distance thresholds between the two mutated positions (A) and between different position types (B). *, FDR < 0.1. **, FDR < 0.01. ***, FDR < 0.001. n.s., non significant.

Figure 5—figure supplement 8. Robustness of enrichments in significant (FDR < 0.2) genetic interactions at different magnitude thresholds in the cis and trans libraries.

(A) Pairs of positions presented as significantly enriched in Figure 5—figure supplement 6A (magnitude threshold of 0.1). Enrichments and their corresponding FDRs were calculated at all 1024 pairs of positions, for each magnitude threshold independently. (B) Enrichments for different position types. Enrichments and their corresponding FDRs were calculated for each magnitude threshold independently. (C) Enrichments for pairs of mutations at difference distance threshold in the structure (pdb: 1fos). Enrichments and their corresponding FDRs were calculated for each magnitude threshold independently.

Figure 5—figure supplement 9. Comparisons of the extent of structural genetic interactions in cis and trans.

(A–B) Proportion of true positive (A) or negative (B) genetic interactions in the original (solid lines) or sub-sampled (open circles) libraries at different p-value cut-offs, showing that the higher prevalence of both positive and negative interactions in the cis library is not the consequence of the different distribution of single mutants effects. Error bars represent 95% confidence intervals around the mean of the 1000 sub-samplings. True interactions are calculated as the total number of interactions called significant at a given p-value threshold multiplied by the true discovery rate (1-FDR) corresponding to that p-value threshold in that library.