Fig 1.
Overview of the proposed method.
Fig 2.
The precision, recall, and F-score of the six methods was compared on synthetic datasets consisting of 15 events sampled from CTMCs parameterized by forests and DAGs.
Fig 3.
The precision, recall, and F-score of six methods was compared on synthetic datasets with 15 events. and uniform noise at six noise level (0.1%, 0.25%, 0.5%, 1%, 2.5%, 5%).
The datasets were generated by sampling from CTMCs parameterized by forests and DAGs. Error bars represent one standard deviation.
Fig 4.
The time cost of one gradient step is plot against different (a) number of accumulated events k and (b) number of events n.
Fig 5.
The oncogenetic graph inferred by TimedHN.
Edge widths and shade are linear to the inter-event hazard rates. The node sizes are linear to spontaneous hazard rates.
Fig 6.
The oncogenetic tree inferred by CAPRESE is shown in subfigure (a) Its edges and nodes are shown in uniform width and size. The Oncogenetic graph inferred by MHN is shown in subfigure (b). Its edge widths and shade are linear to the exponential of inter-event hazard rates. Node sizes are linear to the exponential of spontaneous hazard rates. Solid lines represent edges with positive weights, and dashed lines represent negative ones.