Figure 1.
Variation in phenotypic tests.
(A) Boxplot representation for preprocessed phenotypic data of nine replicates (PA14 wild-type harboring a spontaneous mutation within flgF) for all 136 phenotypic tests. (B) Boxplot representation of preprocessed phenotypic data of all mutants for all 136 phenotypic tests. The low variation of the replicates indicates the robustness of the testing system, whereas the high variation in all mutants reflects the discriminative potential for each test.
Figure 2.
The P. aeruginosa phenome landscape organized by phenotypic similarity determined by 119 phenotypes of 4030 mutants.
We systematically estimated the phenotypic distances between the genotypes, which were the basis for hierarchical clustering and multi-dimensional scaling. (A) Heat map, rows correspond to mutants and columns to different phenotypic tests. (B) Projection onto a two-dimensional grid using multi-dimensional scaling (MDS). Point colors were overlaid by seven cluster groups. (C) MDS data were converted to a contour image using two-dimensional kernel density estimation. Altitudes indicate the density of mutants in the MDS grid. (D) The contour image was depicted as a three-dimensional landscape, in which phenotypically similar mutants are located close to each other, the height indicating the density of mutants in that region of the phenotypic space.
Figure 3.
The phenome landscape for permutated values.
Phenotypic distances between randomly shuffled values served as the basis for hierarchical clustering and multi-dimensional scaling. (A) Heat map, rows correspond to mutants and columns to different phenotype test. (B) Projection onto a two-dimensional grid using multi-dimensional scaling (MDS). Point colors were overlaid by seven cluster groups. (C) MDS was converted to a contour image using two-dimensional kernel density estimation. Altitudes indicate the density of mutants in the MDS grid. (D) The contour image was depicted as a three-dimensional landscape. In comparison to Figure 2 the different cluster group symbols spread homogeneously and in smaller space.
Figure 4.
The global phenotypic network.
The network is based on calculated Jaccard indices (JIs), a measure for gene relatedness. In the largest coherent graph 1196 genes are represented by the nodes and the connections between the genes are drawn for gene pairs with JI>0.51. A zoomable graph with gene names is provided in the supporting information (Figure S1).