Skip to main content
. 2012 Oct 5;137(13):134111. doi: 10.1063/1.4755751

Figure 1.

Figure 1

A simple model demonstrating that BACE correctly deals with poorly sampled transitions, whereas PCCA is confounded by them. This simple model has nine microstates (circles) whose borders are colored (solid black, gray, or dashed black) according to their assignment to three macrostates using either BACE or PCCA. Each microstate has 1000 self-transitions, thick connections represent 100 transitions, medium lines represent 10 transitions, and thin lines represent 1 transition. Therefore, the best coarse-graining into three states is to merge states 0-2, 3-5, and 6-8 because transitions within these groups are fast compared to transitions between the groups. BACE correctly identifies this optimal coarse-graining into three macrostates. However, the poorly sampled transitions between states 2-4 and 5-7 cause PCCA to mistakenly assign states 3-4 with states 0-2 instead of with state 5.