Analysis of temporal change
The most complete time series of ZZB-TE, Cyp6p4-236M andCyp6aap-Dup1 allele frequencies were available from Kabondo, DRC; western Kenya and eastern Uganda. To model allele frequency changes over time we estimated the three parameters of the standard recursive population genetic model (allele frequency at time zero, selection coefficient and dominance coefficient) using a maximum likelihood approach assuming a binomial distribution: an approach previously applied to insecticide target-site resistance mutations [31]. The analysis was performed in R (http://www.r-project.org). An estimated generation time of one calendar month was used as in previous studies [31].