DeepEdit is a fully connected neural network model which takes advantage of the raw electrical signal features flanking the editing sites. A total of 40,823 I-type reads from FY-ADAR2 and randomly chosen 47,757 HFF1 reads were used as the positive and negative controls.