MOESM5 of The ability of single genes vs full genomes to resolve time and space in outbreak analysis

Additional file 5 Maximum likelihood inference of masked sequence location from genomes (left) and GP sequences (right) via a CTMC model implemented in TreeTime. Horizontal bars indicate the posterior distribution of masked tip locations coloured by country (Sierra Leone in blue, Liberia in red, Guinea in green) and location (lighter colours indicate administrative divisions lying towards west of the country). The correct location of each tip is outlined in white with the smaller plot to the right showing only the probability of the correct location. Bars marked with an open circle indicate cases where the correct location is within the 95% credible set and solid circles indicate cases where the location with the most probability is also the correct location. Genomes still perform better in terms of correct guess (0.432 probability that best guess location is true location for genomes versus 0.259 for GP), cross entropy (12012.800 nats for genome versus 24397.109 nats for GP) and mean probability-weighted great circle distance between true location population centroid and estimated location population centroid (87.568 km for genome versus 124.909 km for GP).