Skip to main content

Table 4 Comparison between KLD–RF model and DTI prediction models

From: Random-forest model for drug–target interaction prediction via Kullback–Leibler divergence

 

2D KLD-RF*

3D KLD-RF

Molecular representation

Morgan 2D

E3FP (Omega Conf)

E3FP (Omega Conf)

E3FP (Rdkit Conf)

Number of KLD feature vectors

16

17

17

17

Number of targets

17

17

17 + 1(out-of-set)

17

Out-of-bag score estimate

0.786

0.876

0.874

0.811

Mean accuracy score

0.794

0.882

0.884

0.815

  1. 2D KLD-RF*: Because Sigma opioid receptor of Q3 has 5 ligands not enough to make probability density (2D: 5 ligands vs 3D: 634 conformers), KLD-Q3 feature (KLD feature vector of Q3) was excluded in 2D KLD-RF model and just data of Q3 were included during training/test