A Bayesian approach to incorporate structural data into the mapping of genotype to antigenic phenotype of influenza A(H3N2) viruses
Fig 2
The role of HA positions in antigenic evolution estimated using a structurally naïve model.
(a) Each point represents the posterior inclusion probability for a variable HA1 position. Blue vertical shading indicates positions on the x-axis that are described in antigenic sites A and B (dark blue) and antigenic sites C, D, and E (light blue). (b) Amino acid positions on the surface model of a HA structure are coloured: to the left, by inclusion probability following the colour scheme in a; to the right, antigenic sites A and B are shown in dark blue and sites C, D and E in light blue. (c) Posterior inclusion probability for each variable amino acid position is plotted against distance from the closest residue in a defined antigenic site. Points at 0Å on the x-axis, representing residues in described antigenic sites, have been adjusted in both dimensions to allow visualisation of overlapping points. A red line indicates the mean distance of residues selected with inclusion probability equal to or higher than that on the y-axis. Light red lines show a null distribution derived from 1,000 randomisations retaining the inclusion probabilities from the model fitted to the data. A dashed vertical line at 8Å marks a threshold at which a position is approximately two residues away from an amino acid in a described antigenic site.