Skip to main content

Table 2 Performances of all 9 regression-based Abs_CO prediction methods

From: Protein contact order prediction from primary sequences

 

Correlation Coefficient

Average Percent Correct

Standard Deviation

Method

LR

SVR

NN

LR

SVR

NN

LR

SVR

NN

Method 1 (F3)

0.8571

0.8583

0.8513

0.8630

0.8659

0.8543

0.1099

0.1060

0.1445

Method 1 (F7)

0.8603

0.8543

0.8737

0.8636

0.8681

0.8656

0.1091

0.1094

0.1110

Method 1 (F27)

0.8702

0.8658

0.7713

0.8659

0.8717

0.8086

0.1060

0.1291

1.2222

Method 2 (F3)

0.8477

0.8440

0.8680

0.8577

0.8664

0.8612

0.1179

0.1129

0.1073

Method 2 (F7)

0.8499

0.8455

0.8667

0.8570

0.8669

0.8656

0.1161

0.1142

0.1110

Method 2 (F27)

0.8662

0.8550

0.8233

0.8620

0.8714

0.7016

0.1105

0.1167

3.3827

Method 3 (F3)

0.8375

0.8378

0.8439

0.8542

0.8647

0.8560

0.1200

0.1168

0.1103

Method 3 (F7)

0.8421

0.8381

0.8636

0.8537

0.8637

0.8648

0.1192

0.1165

0.1156

Method 3 (F27)

0.8625

0.8512

0.7706

0.8605

0.8695

0.8106

0.1126

0.1194

0.3662

  1. The correlation coefficients, the average percent correct values, and the standard deviations of the percent correct values for all 9 regression-based Abs_CO prediction methods. The first set of results (rows 3–5) are for Method 1 in which the percentage of residues in alpha helices, p(α), and beta strands, p(β), are used as two factors. In particular, the F3-LR regression formula is Abs_CO = -6.8968p(α) + 7.6216p(β) + 0.0612L + 8.0397. The second set of results (rows 6–8) are for Method 2 in which the numbers of residues in alpha helices, q(α), and beta strands, q(β), are used as two factors. In particular, the F3-LR regression formula is Abs_CO = -0.0591q(α) + 0.0789L + 8.3774. The third set of results (rows 9–11) are for Method 3 in which the numbers of alpha helices, n(α), and beta strands, n(β), are used as two factors. In particular, the F3-LR regression formula is Abs_CO = -0.4184n(α) + 0.1992n(β) + 0.0611L + 8.647.