Skip to main content
. 2023 Dec 20;64(1):42–56. doi: 10.1021/acs.jcim.3c01524

Table 3. Most Common Reaction Encodings for Yield Prediction.

Reaction encoding Short description
Methods developed specifically for the reaction encoding
RXNFP66 Developed to encode SMILES using a pretrained BERT model fine-tuned on Pistachio.
DRFP67 A binary fingerprint based on the symmetric difference of two sets containing the circular molecular n-grams generated from the molecules listed left and right from the reaction arrow in SMILES.
Graph-based encodings Chemprop68,69 implemented the support of CGR63 and uses the pseudomolecule for message passing.
Encodings calculated for the individual components of the reactions
DFT fingerprints Include various features, calculated for each molecule using Quantum Calculation software.
Structural fingerprints (ECFP, Rdkit structural) Fingerprints that are based on the structure of the molecule and calculated structural features,70 use SMILES.
Graph-based encodings AttentiveFP,71 MoGAT.72