Skip to main content

Table 6 Success rate of retrieving a reactant candidate from the PubChem database

From: Substructure-based neural machine translation for retrosynthetic prediction

Tanimoto coefficient

Ratio (%)

No. of discrepant keys

1.00

62

None

\(\ge\).97

70

1

\(\ge\).94

91

2

\(\ge\).90

100

3 or 4

  1. Average length of a molecule in PubChem DB is 42. Test set size = 21,827, PubChemDB size is approximately 154 M