Evolutionary origins of epidemic potential among human RNA viruses

Lu Lu; Liam Brierley; Gail Robertson; Feifei Zhang; Samantha Lycett; Donald Smith; Margo Chase-Topping; Peter Simmonds; Mark Woolhouse

In addition, we applied parsimony reconstruction models to trace trait evolution, using 121 Mesquite V 3.5.1 (http://www.mesquiteproject.org), with the input phylogenies being generated 122 with the protein sequences of each virus genus via maximum likelihood method, using RaxML 123 V 8 (WAG+G, bootstraps n=1000) 12 . We obtained excellent agreement between mean 124 estimates of numbers of forward transitions using parsimony analysis and counts estimated 125 from our discrete traits analysis (Table S1, Fig. S12b) , noting that the parsimony approach does We calculated the depth of the node of each transition on the genus-level phylogenetic tree 151 relative to the ancestral node ( Fig. S4 and Data Files 2 and 3) . We used a beta distributed 152 GLMM (using the R package 'glmmTMB' 13 ) to compare node depth (response variable) across level transition types (explanatory variable) with genus as a random factor. We used a linear 154 mixed model (LMM) with a normal distribution to compare loge-transformed transition rates 155 (response variable) by level transition groups (explanatory variable) with genus as a random 156 factor. We also included genome type/enveloped and vector-borne in the model as potential 157 confounding variables.