Skip to main content

Table 6 Importance of computationally expensive features. A "+" mark indicates if a feature set was included in predictor construction. The feature sets were: AA -26 features directly calculated from the amino acid sequence; PSSM – 22 features as average PSI-BLAST profiles over the input window; PHD – 3 features as average secondary structure prediction scores by the PHDsec predictor; PSI – 3 features as average secondary structure prediction scores by the PSIPRED predictor. F is the total number of features included. The 1st, 5th, and 8th rows correspond to VSL2B, VSL2P, and VSL2, respectively. All accuracies are per-chain accuracies.

From: Length-dependent prediction of protein intrinsic disorder

AA PSSM PHD PSI F SN SP ACC SN S SN L
+     26 77.3 ± 1.1 79.9 ± 0.4 78.6 ± 0.6 75.8 ± 1.3 78.2 ± 1.9
+   +   29 77.9 ± 1.1 80.4 ± 0.4 79.1 ± 0.6 76.6 ± 1.3 78.5 ± 2.0
+    + 29 79.9 ± 1.1 79.9 ± 0.4 79.9 ± 0.5 78.7 ± 1.3 79.9 ± 2.0
+   + + 32 79.8 ± 1.1 80.3 ± 0.4 80.1 ± 0.5 78.6 ± 1.2 79.8 ± 2.0
+ +    48 81.0 ± 1.0 80.4 ± 0.5 80.7 ± 0.6 79.8 ± 1.2 80.9 ± 1.9
+ + +   51 81.3 ± 1.0 81.0 ± 0.5 81.2 ± 0.5 80.0 ± 1.2 81.4 ± 1.8
+ +   + 51 82.3 ± 1.0 80.5 ± 0.5 81.4 ± 0.5 81.1 ± 1.2 81.9 ± 1.9
+ + + + 54 82.3 ± 1.1 81.0 ± 0.5 81.6 ± 0.5 81.3 ± 1.2 82.3 ± 1.8