Figure 1

Illustration of calculating the change trend score sc and the correlation coefficient cc of the match change trend. The two genes X and Y in this example have a positive relationship (here expression levels are not normalized). Only the change trend between time points 5 and 6 is different. The number of matched change after time shifts (by dynamic programming) is less than 7. Thus, the maximal match change trend score sc is 7 in this example with 9 time points. To calculate the correlation coefficient, we first extract the change levels of the matched change trends: $\begin{array}{l}Xm{c}_1^2=1,Xm{c}_2^3=1,Xm{c}_3^4=-1,Xm{c}_4^5=2,Xm{c}_5^6=4,Xm{c}_6^7=-2,Xm{c}_7^8=3\\ Ym{c}_1^2=1,Ym{c}_2^3=1,Ym{c}_3^4=-1,Ym{c}_4^5=2,Ym{c}_5^6=3,Ym{c}_6^7=-3,Ym{c}_7^8=4\end{array}$ The change level between time points 5 and 6 is excluded because the change (inverted) trend is different from the main (positive) change trend. Using the above matched change levels we can obtain a correlation coefficient cc = 0.9597 according to Eq.1.