Fig. 4: CML state-space aligns samples and identifies transition driver genes.

A Using a correlation analysis to identify the DEGs with similar dynamics during CML development, gene modules were identified for the unique DEGs of the three disease states (Es, Ts, Ls; Fig. S3C, S4A). To visualize the expression of each detected gene module that had a mean correlation coefficient > 0.25, the mean log-normalized mean-centered expression of all genes in the module was plotted as function of the CML state-space coordinate for each CML sample. B, C For the early transition point (T-Es; B) and the late transition point (T-Ls; C) DEGs that were involved in state-transition were identified by correlating the expression of each DEG with the shape of the potential around each of the transition points (left top; Fig. S6B). To identify the driver gene processes, the driver genes for each transition point were then extended to their high-confidence interaction partners using STRINGdb (Fig. S6C). The CML contribution were summarized for the resulting protein-protein interaction networks (left bottom). The processes enriched by the STRINGdb extended driver genes involved were identified and their CML contribution was calculated (right).