Antigenic vs genetic distances

[trvrb]

How do the antigenic maps shown compare to BMDS applied to pairwise aa hamming distances alone? How many serious discrpencies that are outside the confidence intervals on map reconstruction? In other words, the authors should demonstrate that the HI data actually provide substantively different results for antigenic grouping than genetic data do alone. This was an important contention of Smith et al (2004), but it was based on the position of only a handful of viruses in that paper. Do the authors have more convincing evidence in this study that the antigenic data are providing substantial extra information beyond the genetic data?

[trvrb]

We have revised the manuscript to include a thorough analysis of the relationship between genetic distance and antigenic distance. This is new figure 1. Rather than constructing BMDS maps with pairwise AA distances, we take a more direct approach and compare pairwise antigenic distance from a BMDS model using only HI data (model 2 in Table 1) and pairwise genetic distances. We examine both pairwise AA distance and pairwise phylogenetic distance, as phylogenetic distance was the basis for our evolutionary diffusion model. We find fairly modest correlations in most cases case (between 0.10 for B/Yam and 0.68 for A/H3N2). Genetic data provides some predictive power for antigenic distance, but is a rather weak predictor alone. This was a big part of our rationale for attempting a joint genetic / antigenic model.