Fig. 4

Performance of various molecular clocks in accommodating the time-dependent rate phenomenon (TDRP). We explored the performance of various clock models in accommodating the TDRP in the context of foamy virus (FV) evolutionary timescale inference under various hypothetical scenarios, where only three nodes are available as calibrating nodes. Two calibration schemes were examined: (i) aggregated-node calibration scheme (top), in which all three calibrating nodes are of similar timescales, and (ii) dispersed-node calibration scheme (bottom), in which calibrating nodes are of different timescales. The aggregated-node calibration scheme was sub-divided into three sub-schemes: (i) shallow-timescale calibration scheme (top left), (ii) intermediate-timescale calibration scheme (top middle), and (iii) deep-timescale calibration scheme (top right). The dispersed-node calibration scheme was also sub-divided into three arbitrary sub-schemes: (i) dispersed-I calibration scheme (bottom left), (ii) dispersed-II calibration scheme (bottom middle), and (iii) dispersed-III calibration scheme (bottom right). Four sets of dates were plotted from different clock implementations and one of our TDRP models: (i) simple power rate decay model (green), (ii) strict clock model (yellow), (iii) log-normal relaxed clock model (orange), and (iv) random-local relaxed clock model (red). FV evolutionary timescales inferred from the host timescales (reference timescales) are in white. Horizontal dotted grey lines are median reference timescales. The results are plotted on a log scale, and arranged in such a way that the evolutionary timescales of shallow nodes are on the left, and those of deep nodes are on the right. The node numbers refer to those in Fig. 1a. Calibrating nodes are indicated by asterisks ‘*’. The summary of the results can be found in Additional file 1: Table S5 & S6