Figure 1.
The heatmap shows amino acid frequencies taken from 2032 bacterial genomes. Light colors represent higher frequencies while darker colors represent lower frequencies. The red and blue colors on the top bar represent GC content, where dark red and blue indicates AT- and GC-rich genomes, respectively. Genomes having GC/AT content close to 50% are represented by lighter grey colors. The bottom bar shows colors designating each genome’s phylum, which are detailed in the figure.
Figure 2.
Principal component analysis plot.
The plot shows two principal components resulting from a principal component analysis performed on the amino acid frequencies taken from 2032 bacterial genomes. The first principal component (PC1) was strongly associated with genomic %AT (decreasing left to right), while the second principal component (PC2) was associated with phyla.
Figure 3.
The heatmap shows codon frequencies from 2032 bacterial genomes. Light colors represent higher frequencies while darker colors represent lower frequencies. The red and blue colors on the top bar indicate GC content, where dark red and blue represents AT- and GC-rich genomes, respectively. Genomes having GC/AT content close to 50% are represented by lighter grey colors on the top bar. The bottom bar shows colors indicating each genome’s phylum, which are described in the figure.
Figure 4.
Amino acid usage bias versus codon usage bias.
The Figure shows a GAM regression with amino acid usage bias on the y-axis (AAUB) as response regressed against the smooth of codon usage bias (CUB) on the x-axis. The dots represent the residuals together with the smoothed regression line. Both left- and right panels represent the same model, but the right panel is based on a GAMM model where strain, genus and phylum, with respect to AT content, are included as hierarchical random effects.
Figure 5.
Codon and amino acid usage bias versus genomic %AT.
The panels show codon and amino acid usage bias (vertical axis, left and right panel, respectively) plotted against genomic fraction of AT (horizontal axis) for 2032 genomes. The blue line shows what would be expected if the codon and amino acid usage bias were perfectly symmetrical for AT and GC-rich genomes.
Figure 6.
GAMM regression model of KL against AT content and AAUB.
The panels show a GAMM regression model with relative entropy (KL) as response with genomic %AT and amino acid usage bias as predictors (left and right panels, respectively). Strain, genus and phylum have additionally been included as random effects with respect to genomic %AT. The dots represent the model residuals with respect to each predictor (AT content and AAUB) together with the spline estimated regression line. The shaded area surrounding the regression line indicates an interval of two standard errors.