Adding new specific reactions
Hello,
-
If we want to add new reactions and train a new model, can we add some with a format different than SMILES ?
-
And do you have any advice when we add new reactions? Some points we should pay attention to?
-
What format do you have in mind? We have developed with SMILES in mind, and I don't know if any other language for describing molecules like SELFIES are supported by e.g. RDKit or RDChiral which are hard dependencies for extracting templates
-
We have experience with adding several different sources like Reaxys, Pistachio and AstraZeneca ELNs, and I don't think any of them are inherently different than USPTO. Reaxys for instance much more varied reaction space, with a lot of unclassified reactions by NextMove - but the pipelines can handle this.