Skip to main content

Table 2 Descriptors used to train VAEs for different downstream problems

From: Improving VAE based molecular representations for compound property prediction

Dataset

Selected descriptors

Pearson correlation coefficient (r)

Aqueous solubility (logS)

MolLogP

− 0.8

PEOE_VSA6

− 0.61

MolWt

− 0.59

Molecular lipophilicity (logD)

MolLogP

0.43

NumAromaticRings

0.29

RingCount

0.26

Blood–brain barrier penetration (logBB)

TPSA

− 0.57

NumHAcceptors

− 0.43

NumHeteroatoms

− 0.43