Skip to main content

Table 3 Results under 5CV of different feature combinations considered on RPI1807 and NPInter v2.0

From: EDLMFC: an ensemble deep learning framework with multi-scale features combination for ncRNA–protein interaction prediction

Dataset Combinations of features ACC (%) TPR (%) TNR (%) PPV (%) F1 (%) MCC (%) AUC (%)
RPI1807 Sequence 92.1 ± 1.3 94.5 ± 2.7 85.1 ± 2.9 94.9 ± 0.8 94.7 ± 0.9 79.5 ± 2.9 96.2 ± 0.8
  Secondary structure 92.8 ± 1.2 96.6 ± 1.9 81.5 ± 6.6 94.0 ± 2.1 95.2 ± 0.7 80.7 ± 3.5 96.1 ± 1.1
  Tertiary structure 72.9 ± 9.6 85.7 ± 21.7 28.7 ± 28.7 82.9 ± 6.9 79.7 ± 12.7 26.9 ± 21.5 81.6 ± 8.7
  Sequence + secondary structure 93.8 ± 0.6 96.6 ± 1.2 85.5 ± 2.5 95.2 ± 0.8 95.9 ± 0.4 83.5 ± 1.5 96.5 ± 0.8
  Sequence + tertiary structure 92.6 ± 1.6 95.1 ± 2.2 85.1 ± 2.9 95.0 ± 0.9 95.0 ± 1.1 80.5 ± 3.9 96.3 ± 0.8
  Secondary structure + tertiary structure 92.4 ± 0.4 96.5 ± 1.5 80.6 ± 5.6 93.7 ± 1.6 95.0 ± 0.2 79.7 ± 1.3 96.2 ± 0.9
  All features 94.3 ± 0.2 97.4 ± 1.0 85.1 ± 1.9 95.1 ± 0.6 96.2 ± 0.2 84.7 ± 0.7 96.7 ± 0.8
NPInter v2.0 Sequence 87.7 ± 0.8 89.7 ± 1.1 85.7 ± 2.4 86.3 ± 1.9 87.9 ± 0.7 75.5 ± 1.6 94.6 ± 0.3
  Secondary structure 78.8 ± 1.4 87.5 ± 1.4 70.1 ± 2.4 74.6 ± 1.6 80.5 ± 1.2 58.5 ± 2.7 88.1 ± 1.0
  Tertiary structure 54.7 ± 3.8 68.1 ± 27.0 41.4 ± 33.6 58.7 ± 8.7 56.3 ± 10.6 10.9 ± 8.7 59.9 ± 2.7
  Sequence + secondary structure 89.1 ± 0.9 91.2 ± 1.1 86.9 ± 1.5 87.5 ± 1.3 89.3 ± 0.8 78.3 ± 1.7 95.4 ± 0.3
  Sequence + tertiary structure 88.9 ± 0.8 91.1 ± 1.1 86.8 ± 1.3 87.3 ± 1.1 89.2 ± 0.7 77.9 ± 1.5 95.2 ± 0.5
  Secondary structure + tertiary structure 83.5 ± 1.0 88.6 ± 1.5 78.5 ± 2.6 80.5 ± 1.8 84.3 ± 0.7 67.5 ± 1.9 92.0 ± 0.4
  All features 90.0 ± 0.7 92.2 ± 1.1 87.6 ± 0.9 88.2 ± 0.8 90.2 ± 0.7 80.0 ± 1.4 96.2 ± 0.3
  1. The values in bold indicate this performance metric is the best among the three methods
  2. The mathematical notation (±) represents standard deviation
\