Fig. 1

Outline of DMRscaler method. A Flowchart of decision tree for DMRscaler. Starts with specification of a window size vector that lists the window sizes to use for identifying DMRs in each layer, and the coordinates of CpG probes with precomputed p-values for each CpG. Additionally Type I error control should be done using these precomputed p-values by setting a p-value cutoff value to ensure that identified DMRs have the expected Type I error rate with respect to their association with the condition of interest. Window significance is determined using Eq. (1). Adjacent significant windows are merged forming the Next_Layer. For the first iteration, the returned Layer_1 is set to this Next_Layer, for subsequent iterations the returned Layer_i is set to the result of integration of Next_Layer with Prev_Layer. Integration of layers is described in 1C. Prev_Layer is updated to Layer_i before proceeding to iteration i + 1. After the largest window size layer is generated, a list is returned of the results from each iteration of the algorithm. B Graphical description of algorithm. At the top shows representation of CpGs ordered by position and associated with a significance value. Windows are laid over the ordered CpGs and selected if the window score is significant. Adjacent windows are then merged. If a Prev_layer has been assigned, then integration occurs. C Integration procedure. For each Next_Layer DMR, all overlapping Prev_Layer DMRs are identified. A subtraction set is generated by individually subtracting each overlapping Prev_Layer DMR from the Next_Layer DMR. Subtraction involves removing overlapping CpGs from the Next_Layer. If all elements of the subtraction set are significant when rescored with the window scoring function, then the Prev_Layer and Next_Layer regions are merged in the Integrated_Layer, otherwise the Prev_Layer DMRs are used in the Integrated_Layer. This procedure ensures that the broader Next_Layer DMRs are only included if no single Prev_Layer DMR was responsible for the significance of the region identified in the Next_Layer