Fig. 3

Effect of sampling on the phylogenetic tree. a Representation of a transmission tree, where nodes represent individuals, numbers represent the order of infection chronologically and the arrows show the direction of transmission. b Phylogenetic tree when all individuals in the outbreak are sampled. Transmission pairs are not necessarily paired on the tree as they may not be the most similar within the context of the outbreak. For example, if we assume that 1 had a long, chronic TB infection then because of the amount of diversity that can accumulate over time it is possible for the genomes from 2 and 3 to be more closely related to each other than to the genome from 1, even though 1 infected them both. This is because the strain that was sampled from 1 has evolved since 1 infected 2 and 3. While rejecting pairs not adjacent on the phylogenetic tree seems sound when sampling is sparse (as transmission pairs would then be relatively rare in the dataset and closer in phylogenetic distance than typical pairs of tips), when sampling is dense (as is desirable in epidemiological investigations). c Individuals 2, 3, 4 and 8 have not been sampled for the reconstruction of this tree. This makes the distances between the average pair of tips in the tree larger, highlights the close phylogenetic distance between 6 and 7 and (correctly) suggests transmission occurred between these individuals