Skip to main content

Table 4 Comparison of different combinations of features in 10-fold cross validation

From: Predicting protein-binding regions in RNA using nucleotide profiles and compositions

 

Sensitivity

Specificity

Accuracy

PPV

NPV

MCC

mPWM

89.09%

90.60%

89.87%

89.67%

90.06%

0.797

dPWM

90.48%

92.06%

91.31%

91.27%

91.34%

0.826

compositions

71.44%

88.23%

80.20%

84.76%

77.12%

0.608

mPWM + dPWM

91.46%

91.98%

91.73%

91.27%

92.16%

0.834

mPWM + compositions

91.31%

91.55%

91.43%

90.83%

92.00%

0.828

dPWM + compositions

91.07%

92.53%

91.83%

91.78%

91.88%

0.836

mPWM + dPWM + compositions

91.61%

92.39%

92.02%

91.69%

92.31%

0.840

  1. Using all 3 features showed the best performance. mPWM: mono-nucleotide position weight matrix, dPWM: di-nucleotide position weight matrix, compositions: frequency of mono-nucleotides, di-nucleotides, and tri-nucleotides in the RNA sequence