Skip to main content

Table 3 Performance comparison of proposed features with AAC, DPC, PSSM, and the combined features with highest performance values for each class highlighted in bold

From: TNFPred: identifying tumor necrosis factors using hybrid features based on word embeddings

  Cross-validation data
Feature types Acc (%) Spec (%) Sen (%) MCC
AAC 60.69 ± 12.13 59.61 ± 15.82 68.01 ± 19.77 0.24 ± 0.09
DPC 70.42 ± 13.97 72.89 ± 17.91 51.67 ± 20.14 0.24 ± 0.13
AAC-DPC 86.48 ± 5.82 88.83 ± 8.04 69.34 ± 16.62 0.53 ± 0.08
PSSM 89.57 ± 6.68 91.88 ± 5.73 73.34 ± 18.33 0.61 ± 0.16
PSSM-AAC 91.17 ± 3.18 93.19 ± 4.11 76.34 ± 12.91 0.66 ± 0.1
PSSM-DPC 91.25 ± 3.29 93.85 ± 4.58 72.34 ± 12.58 0.64 ± 0.09
PSSM-DPC-AAC 91.25 ± 2.84 93.67 ± 3.85 73.67 ± 14.45 0.64 ± 0.1
Proposed features 95.82 ± 1.67 97.59 ± 2.15 83.67 ± 7.45 0.83 ± 0.06
  Independent data
Feature types Acc (%) Spec (%) Sen (%) MCC
AAC 46.92 ± 25.59 42.37 ± 33.15 81.25 ± 34.49 0.20 ± 0.07
DPC 82.05 ± 23.71 84.63 ± 29.06 62.75 ± 39.92 0.46 ± 0.28
AAC-DPC 93.65 ± 2.66 94.95 ± 3.26 83.75 ± 9.59 0.73 ± 0.09
PSSM 94.85 ± 2.82 97.56 ± 2.85 74.5 ± 28.35 0.72 ± 0.26
PSSM-AAC 95.77 ± 1.45 97.42 ± 2.14 83.25 ± 11.31 0.81 ± 0.06
PSSM-DPC 95.94 ± 1.33 97.81 ± 1.49 82 ± 8.8 0.81 ± 0.07
PSSM-DPC-AAC 95.14 ± 2.02 96.77 ± 2.48 83 ± 9.7 0.78 ± 0.08
Proposed features 96.49 ± 4.34 98 ± 5.27 85 ± 17.48 0.86 ± 0.13
  1. (Each result is reported in format: m ± d, where m is the mean and d is the standard deviation across the ten runs)