Skip to main content

Table 4 Accuracy comparison of PROTNN with other classification techniques

From: ProtNN: fast and accurate protein 3D-structure classification in structural and topological space

Dataset Classification approach
  Blast Sheba FatCat CE LPGBCMP D&D GAIA ProtNN ProtNN*
DS1 0.88 0.81 1 0.45 0.88 0.93 1 0.97 0.97
DS2 0.82 0.86 0.89 0.49 0.73 0.76 0.66 0.8 0.89
DS3 0.9 0.95 0.84 0.59 0.90 0.96 0.89 0.96 0.97
DS4 0.76 0.92 1 0.46 0.9 0.93 0.89 0.97 0.97
DS5 0.86 0.99 0.94 0.76 0.87 0.89 0.72 0.9 0.94
DS6 0.78 1 0.94 0.81 0.91 0.95 0.87 0.96 0.96
Avg. accuracy1 0.83 ±0.05 0.92 ±0.07 0.94 ±0.06 0.59 ±0.15 0.86 ±0.06 0.9 ±0.07 0.84 ±0.12 0.93 ±0.06 0.95 ±0.03
Avg. distances2 0.14 ±0.07 0.05 ±0.07 0.04 ±0.05 0.38 ±0.15 0.11 ±0.03 0.7 ±0.04 0.14 ±0.09 0.05 ±0.03 0.02 ±0.01
Rank 8 4 2 9 6 5 7 3 1
  1. 1Average classification accuracy of each classification approach over the six datasets
  2. 2Average of the distances between the accuracy of each approach and the best obtained accuracy with each dataset
  3. The boldface numbers highlight the best performance