Skip to main content

Table 2 Results of testing our model and DeepBind on RNA sequences of 25 nucleotides. catRAPID could not be tested on RNA sequences of 25 nucleotides since the minimum length of an RNA sequence required by catRAPID is 50 nucleotides

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

 

#RBP-binding

      

RBP

RNA regions

Sensitivity

Specificity

Accuracy

PPV

NPV

MCC

Our model

       

FUS

64

93.75%

94.00%

93.90%

90.91%

95.92%

0.873

FXR1

67

97.01%

94.00%

95.21%

91.55%

97.92%

0.902

FXR2

80

66.25%

94.00%

81.67%

89.83%

77.69%

0.638

IGF2BP2

79

74.68%

94.00%

85.47%

90.77%

82.46%

0.709

LIN28A

82

85.37%

94.00%

90.11%

92.11%

88.68%

0.801

QKI

77

84.42%

94.00%

89.83%

91.55%

88.68%

0.793

TARDBP

94

12.77%

94.00%

54.64%

66.67%

53.41%

0.117

Weighted average

 

70.72%

94.00%

83.83%

90.14%

80.54%

0.676

DeepBind

       

FUS

64

32.81%

42.00%

38.41%

26.58%

49.41%

-0.246

FXR1

67

11.94%

44.00%

31.14%

12.50%

42.72%

-0.444

FXR2

80

15.00%

55.00%

37.22%

21.05%

44.72%

-0.320

IGF2BP2

79

41.77%

51.00%

46.93%

40.24%

42.58%

-0.072

LIN28A

82

12.20%

52.00%

34.07%

17.24%

41.94%

-0.382

QKI

77

83.12%

75.00%

78.53%

71.91%

85.23%

0.576

TARDBP

94

52.13%

92.00%

72.68%

85.96%

67.15%

0.484

Weighted average

 

36.28%

58.71%

48.91%

40.53%

54.29%

–0.051

  1. The specificity of our method is the same for all RBPs because it used a same set of negative data for all RBPs with a single model, whereas DeepBind has distinct models for each RBP