MERS-CoV recombination: implications about the reservoir and potential for adaptation

Dudas, Gytis; Rambaut, Andrew

. Trees recovered using GARD (Kosakovsky Pond et al., 2006) across the breakpoint at position 23722. NJ trees reconstructed by GARD across the first identified breakpoint. Tree from positions 1-23722 on the left and positions 23723-30126 on the right. The same tips in both trees are connected by coloured lines to indicate phylogenetic incongruity. Figure S2 . Empirical rate heterogeneity in MERS-CoV genome. Posterior estimates of the ratio between the molecular clock rates estimated independently from GARD-inferred fragment 2 (positions 23723-30126) and fragment 1 (positions 1-23722) under independent or linked tree models derived from 3 independent marginal likelihood analyses. Dotted lines indicate the mean of the distribution and numbers next to the line show the median and the 95% highest posterior density intervals. . Window-based estimates of recombination rate. Inferred recombination rates for 300 nucleotide-long windows in MERS-CoV genome (top), πBUSS-simulated sequences with 1.3× rate heterogeneity (middle) and 3× rate heterogeneity (bottom) under a nucleotide substitution model. Recombination rates that are above the inferred genome-wide recombination rate are in red. Simulated rate heterogeneity is sufficient to mislead this method, although the inferred recombination rates in the last third of the MERS-CoV genome are much greater than those inferred from the simulated data. Figure S6 . Window-based estimates of polymorphic site density. Inferred polymorphic site densities for 300 nucleotide-long windows in MERS-CoV genome (top), πBUSS-simulated sequences with 1.3× rate heterogeneity (middle) and 3× rate heterogeneity (bottom) under a nucleotide substitution model. Windows are coloured red if their recombination rate is above the inferred genome-wide recombination rate. Extreme rate heterogeneity (3×) results in a higher density of polymorphic sites in the region with the higher rate. Figure S7 . Homoplasy degrees inferred by BEAST (Drummond et al., 2012) . Position along the genome is shown on the x axis and homoplasy degree, the number of times a particular mutation has occured in excess in the tree, is shown on the y axis. Individual mutations are marked by vertical lines, synonymous ones in green and non-synonymous in red with transparency representing the posterior probability of a given homoplasy degree for each mutation. The ratio of apparent homoplasy over synapomorphy kernel density estimates (bandwidth=0.1) is shown in blue for synonymous (dashed) and non-synonymous (solid) sites separately. Arrows at the top indicate the positions and names of coding sequences within the MERS-CoV genome. Figure S8 . Host association indices for variable sites. Estimates for the association between particular alleles and host. The association index is an adapted version of the χ 2 df statistic of LD (Hedrick and Thomson, 1986) , and quantifies how well one can predict the allele at any given polymorphic site, given the host it was isolated from. No perfect associations (association index = 1.0) between particular alleles and host (human or camel) were found. Figure S9 . Maximum likelihood phylogenies across MERS-CoV genome. Maximum likelihood phylogenies recovered with PhyML (Guindon and Gascuel, 2003) under GTR+Γ 4 (Tavaré, 1986; Yang, 1994) nucleotide substitution model across 4000 nucleotide fragments derived from the MERS-CoV genome. Each tip is connected to its counterpart in phylogenies of neighboring fragments and coloured se