This model is intended for retrosynthetic prediction of the reagents required to obtain a given product. It is based on a transformer architecture with a Multi-head Attention mechanism contains 6 layers and 8 “heads”. The Atom-in-SMILES representation of molecules was used as input files. The model was trained on the USPTO dataset and oligopyrroles synthesized in our laboratory.

This model is based on an RFR/ECFP for predicting the yield of condensation reactions between pyrroles and dipyrromethanes with aldehydes.