Fig. 3

Detecting the critical transitions for two complex diseases. The HMM-based method was applied to identify the pre-disease states for two disease datasets. a The Consistence score based on the top differential-differential genes form each candidate transition time point for HCC. It can be seen that the most significant signal appears at the 4-th sampling stage (very early HCC). b The Consistence score based on the top differential-differential genes form each candidate transition time point for live influenza infection. It can be seen that the most significant signal appears at the 29 hours and 36 hours, which is in coincident with the clinical diagnosis. c We illustrate the clustering result respectively for genes at “very early HCC" stage (orange bubbles) and “low-grade dysplastic" stage (blue ones) for HCC. d We illustrate the clustering result respectively for live influenza infection at 29 h (orange bubbles) and 5 h (blue ones). Clearly, comparing with the control group, the P-value of each gene group is more significant at the critical stage than that at the early stage. The top differential-expression and fold-change gene groups were selected to proceed with functional analysis