Skip to main content

Table 5 The comparison of different methods for the average evaluation values in all cells

From: LightCpG: a multi-view CpG sites detection on single-cell whole genome sequence data

Dataset Methods Acc(%) AUC Fscore(%) MCC(%) SP SE(%)
HCCs LightCpG 92.06 0.9616 84.66 78.97 93.73 86.84
  DeepCpG 92.34 0.9689 86.42 81.24 92.95 90.59
  RF Zhang 88.41 0.9351 79.93 72.08 89.38 85.59
HepG2 LightCpG 83.20 0.9213 81.73 67.36 89.96 78.32
  DeepCpG 84.17 0.9248 82.52 68.22 85.27 83.40
  RF Zhang 81.16 0.8942 80.17 63.20 87.39 76.29
  1. 1LightCpG employs three types of features (sequence feature, structural feature and positional feature) and LightGBM [36] to identify the CpG sites
  2. 2DeepCpG [35] embodies the connection between various cells by using the deep learning model Gated Recurrent Network (GRU) and also extracts features from the DNA sequence by the convolutional neural network (CNN) and one additional fully connected hidden layer. Then DeepCpG uses Fully Connected Neural Network to identify CpG sites
  3. 3RF Zhang [1] extracts the genomic positional features, neighbor features, sequence properties and sic-regulatory elements to identify the CpG sites
  4. 4The boldface is the best value in the column