Select combination and strain

Number of strong interactions per drug

Strong synergies and antagonisms

For the drugs of the selected combinations, strong interactions are shown (FDR < 0.05 and effect size > |0.1|). The effect size is quantified as absolute value of first or third quartile for synergy and antagonism, respectively, of the distribution of Bliss scores for the selected combination (Methods).

Growth curves

Optical density at 595 nm (OD595nm) was monitored over time until cells reached stationary phase (Methods). OD595nm values are shown after background subtraction, before normalizing by no-drug controls (Methods). The arrangement of the plots is the same as in the corresponding checkerboard plot (see “Checkerboards”), with the curves arrayed to form 4x4 concentration matrices for each drug combination, including single-drug controls (left column for donor and bottom row for recipient drugs) and no-drug controls. The three concentrations at which each drug was tested – expressed in µg/ml - are annotated together with drug names in each single-drug column or row. Experiments can be selected according to each drug testing mode (as donor or recipient – see Methods), plate batch (corresponding to recipient biological replicate), donor drug biological replicate. Note that for single-donor controls (left column), growth curves from three single-donor control wells in the selected plate are shown; for single-recipient controls, growth curves from the recipient control plates included in the selected batch are shown (Methods). Each subplot shows two growth curves from technical replicate wells.

Checkerboard viewer

4x4 checkerboards assembled as explained in “Growth curves”. Shown values correspond to optical density at 595 nm (OD595nm) measured at the transition between exponential and stationary phase (8h for S. aureus, 5.5 h for B. subtilis and 3.7 h for S. pneumoniae). This value was then normalized by the robust mean of corresponding values from six no-drug controls included in each plate (Methods). Each subplot indicates which drug of the pair was tested as donor, the single-donor biological replicate, the recipient plate batch used and which technical replicate of the plate is shown. Concentrations – expressed in µg/ml - are shown next to drug names for each subplot. Significance and effect size values are shown for each combination. Adjusted p-value 'padj' is based on a non-parametric (resampling) test and 'efsize' is the upper or lower quartile of all the Bliss scores observed for this combination (Methods).

Interaction score matrix

For each combination, all Bliss interaction scores observed in the screen are shown. Negative Bliss scores indicate synergy, positive scores indicate antagonism. In rows we enumerate concentration ratios of the combination. Concentrations are expressed in µg/ml. Columns follow the notation: (plate batch, donor biological replicate, technical replicate well, drug tested as donor). Grey cells indicate that the concentration ratio was not tested in that particular [batch - biological replicate - donor mode]. The cumulative effect size of the combination is shown on top of the heatmap, as first quartile value of the interaction score distribution for synergies, and third quartile for antagonisms (Methods).

Drug annotation

The table shows manually curated drug annotation (Supplementary Tables 1 and 6). The column “Main or non-antibiotic screen” specifies whether drugs are part of the main antibiotic screen in S. aureus, S. pneumoniae and B. subtilis or of the non-antibiotic screen performed in S. aureus DSM 20231. Drugs of the selected combination are highlighted in orange.

Prevalence of interactions by drug class

Interactions between drug classes are visualized in the network plot. The thickness of edges represents the number of interactions between the drug classes; yellow and green edges correspond to antagonisms and synergies, respectively. The exact numbers for each class-class edge are shown in the table on the right. Different thresholds can be set according to the minimum number of interactions for each class-class link and interaction effect size.

Explore pairwise class-class interactions

Specify classes for which you want to view the interactions. The table will display only interacting combination pairs with one of the drugs of class 1 and the other from class 2

Conservation across Gram-positive strains

Select strains to compare drug-drug interactions. For S. aureus, for which two strains were tested, we differentiate in “strong and conserved”, “weak and conserved” and “strain-specific” interactions. Interactions are strong and conserved if in both strains the interaction is significant (FDR < 0.05) and the effect size exceeds |0.1|. Weak and conserved interactions are significant only in one of the two strains and detected in the other strain based on a lower effect-size threshold (|0.08|). Strain-specific interactions are strong interactions of opposing signs in the two strains. Non-conserved interactions respect strong significance and effect-size cutoffs only in one strain and do not reach the weak effect-size in the other strain (Methods). When two species are compared, we only distinguish between conserved (respecting strong significance and effect-size cutoffs in both species) and non-conserved interactions (respecting strong significance and effect-size cutoffs only in one species).

Effect size comparison

Interactive scatter plot of effect size (Bliss score quartiles) in the selected strain pair allows to compare the weak and strong interactions. Hover over a point to view the combination ID. Note that non-conserved interactions in this scatter plot and in the bar chart above include strain specific interactions.

Conservation across all Gram-positive strains

For each pairwise combination, a single effect size value was derived from the distribution of interaction scores, consisting of at least 72 interaction scores including all replicates of individual concentration combinations for each drug pair. The first and third quartile values of this distribution were taken as effect size values (bliss_score) for synergies and antagonisms, respectively, with negative values corresponding to synergies and positive values to antagonisms. P-values (p_adjusted) were Benjamini-Hochberg adjusted for multiple testing.
Download full table