Skip to main content

Table 1 Performance results from implemented human DLPR models

From: Supervised promoter recognition: a benchmark framework

   CNNProm dataset ICNNP dataset
   1426 promoters 19,811 promoters 21,237 promoters 7156 promoters
   8256 non-promoters 27,731 non-promoters 35,987 non-promoters 5235 non-promoters
   TATA Non-TATA Complete dataset Complete dataset
   MCC PPV Sn Sp MCC PPV Sn Sp MCC PPV Sn Sp MCC PPV Sn Sp
OS\(^*\) 1 90 95 98 89 90 98 82 79
2 90 87
3
CT\(^*\) 1 92 95 91 99 72 85 83 89 73 85 82 91
CI 90–93 94–96 89–93 99–99 71–73 81–88 80–86 86–92 72–74 82–87 79–84 89–93
2 64 83 88 74
CI 61–67 80–87 83–93 66–82
3
CI
CT\(^-\) 1 86 82 94 97 73 91 76 95 76 88 82 94 69 76 87 84
2 69 76 87 84 70 88 87 83
3 16 18 91 28 − 5 41 93 05 − 4 37 95 03 07 59 96 07
CT\(^+\) 1 94 93 97 99 76 89 83 93 80 90 84 95 74 93 84 91
2 72 79 86 87 82 95 89 93
3 19 18 99 22 − 3 41 96 03 0 37 97 03 07 59 96 07
   DProm dataset
   3065 promoters 26,532 promoters 29,597 promoters
   3065 non-promoters 26,532 non-promoters 29,597 non-promoters
   TATA Non-TATA Complete dataset
   MCC PPV Sn Sp MCC PPV Sn Sp MCC PPV Sn Sp
OS* 1 62 75 91 26 58 83
2
3 88 93 95 92 97 95
CT* 1
CI
2
CI
3 90 95 95 95 91 96 95 96 91 97 94 97
CI 88–92 92–98 93–97 91–98 90–92 94–97 94–97 94–97 91–92 95–98 92–96 95–98
CT 1 16 18 91 28 − 5 41 93 05 − 4 37 95 03
2 27 58 87 37
3 86 95 91 95 91 98 93 98 91 96 95 96
CT+ 1 65 74 95 67 14 54 82 29 12 53 84 26
2 27 58 86 38
3 99 1 99 1 94 98 96 98 94 98 96 98
  1. OS indicates model results from original studies. CT shows cross-testing results from our implemented models. (1) CNNProm, (2) ICNNP, (3) DProm model. CT\(^+\): results from training and testing datasets with some overlapping sequences. CT\(^-\): results from properly split training and testing datasets having no overlapping sequences. CT\(^*\) and OS\(^*\): results obtained from 10-fold cross-validation. All values expressed as percentages. \(99\%\) confidence interval values for our 10-fold cross validation results shown within CI rows