Table 1 Summary of the results for the typical AIBL ordinary least squares approach.

From: Aqueous pKa prediction for tautomerizable compounds using equilibrium bond lengths

Metric

O–H (i)

C–O (ii)

C=C (iii)

C–C (iv)

C=O (v)

Slope (+/−)

−

+

×

×

×

r2 (train)

0.56

0.72

0.38

0.15

0.38

MAE (7-fold CV) (train)

0.60

0.41

0.65

0.88

0.73

RMSEE (7-fold CV) (train)

0.75

0.57

0.89

1.10

0.90

MAE (test)

0.31

0.24

0.43

0.67

0.56

RMSEP (test)

0.41

0.34

0.58

0.86

0.69

s.d. (test)

0.28

0.24

0.40

0.55

0.41

r2 obs vs pred (test)

0.90

0.92

0.69

0.66

0.20

  1. (Upper) Statistics for the single bond length models obtained via ordinary least squares regression. The row labelled “slope” features a “+” sign for a positive slope (i.e. pKa increases with increasing bond distance), and a “−” sign to denote a negative slope (i.e. pKa decreases with increasing bond distance). The squared correlation coefficient was not significant enough (“×”) to assign a slope direction for iiii, iv and v.
Back to article page