Accounting for sampling bias in ancestral state reconstruction

Viral transmission plays an essential role in our understanding of and response to infectious diseases, for example, by informing policy decisions about transportation and borders. Phylogenetic methods take advantage of the evolutionary relationships between the genome sequences of viruses to infer geographical locations of unobserved ancestors from sampled data. Here I introduce a new approach to examine the inference of ancestral locations and predict the geographic movement of viral lineages from the known locations of samples. In contrast to existing methods, my method accounts for differences in sampling policies among areas, to avoid biased inference of the ancestral locations. I begin by summarizing existing methods for ancestral state reconstruction. I then introduce an ancestral state reconstruction method that accounts for variation in sampling rate among locations and compare it to the classic Maximum Likelihood method. I show that my method infers ancestral states for small trees more accurately than this classic approach.