Fig. 2

Overall flowchart of ICGEC. The core ICGEC algorithm begins with two normalized matrices (\( G2{M}_{G\ast M}^{C_{1m}} \), \( G2{M}_{G\ast M}^{C_{2m}} \)) from cell lines C₁ and C₂ to be compared. Then, the w_m-based weighted correlation for each gene pair is calculated to produce the context matrices of the genes \( G2{G}_{G\ast G}^{C_1} \) and \( G2{G}_{G\ast G}^{C_2} \), respectively. Next, the program enters into the 1st inner loop, where w_g is iteratively updated until convergence. Then, the w_g-based weighted correlation for each mark pair is calculated to produce the context matrices of the marks (\( M2{M}_{M\ast M}^{C_1} \), \( M2{M}_{M\ast M}^{C_2} \)) from the corresponding normalized matrices \( G2{M}_{G\ast M}^{C_{1g}} \) and \( G2{M}_{G\ast M}^{C_{2g}} \), respectively. Once it is done, the program enters into the 2nd inner loop, where w_m is iteratively updated until convergence. Finally, the program judges whether the new weights are sufficiently close to the weights in the last round. If yes, the program ends up with w_g and w_c returned as the gene score and mark score, respectively; otherwise, the program enters the outer loop. It should be noted that all of the produced weighted correlations of less than zero must reset to 0 during the iterative process to meet the demands of biological significance. w_g represents gene weights, and w_m represents mark weights. The right panel highlights the calculation process for the two inner loops