Fig. 1

Feature generation overview. (a) Methylation level of sample 1 (S1) and sample 2 (S2) for a DMR from the training set. The overlapping fixed size window is used around individual cytosine (C) in the DMR for feature extraction. (b) Extracted features: p-value and difference in methylation level for each CG site. (c) Histogram of scores computed from the extracted features and (d) histogram of normalized scores. (e) Methylation level of S1 and S2 for a non-DMR from the training dataset. The overlapping fixed size window is used around individual C in the DMR for feature extraction. (f) Extracted features: p-value and difference in methylation level for each CG. (g) Histogram of scores computed from the extracted features and (H) histogram of normalized scores. (i) Mean and standard deviation of histogram features for complete training data for DMRs (blue) and non-DMRs (pink). (j) Testing and DMR prediction on new dataset