[Data augmentation in a TTL] - Fictive dataset (1M) // 100 reactions per reaction template (13'953 template extracted from USPTO-FULL IBM version)

Equilibrated version of the fictive dataset, containing 1'097k reactions with up to 100 reactions per radius 1 reaction template (13'953 reaction templates from USPTO-full, IBM version). Title of the manuscript: "Data augmentation in a Triple Transformer Loop retrosynthesis model" Abstract: Reactions in the US Patent Office (USPTO) are biased towards a few over-represented reaction types, which potentially limits its usefulness for computer-assisted synthesis planning (CASP). To obtain an equilibrated dataset, we applied retrosynthesis templates to USPTO molecules as products (P) to generate starting materials (SM). We then used transformer T2 from our recently reported triple transformer loop (TTL) retrosynthesis model to predict reagents (R) for the SM®P reaction. Finally, we validated the prediction by requesting a high confidence prediction (>95%) for the prediction of P from SM+R by TTL transformer T3. We generated up to 5,000 reactions per template, resulting in 27.5 million validated fictive reactions covering the chemical space of the original UPSTO dataset. To exemplify the use of this dataset, we show that a single-step retrosynthesis transformer model trained with a template equilibrated subset of 1,097,374 fictive reactions outperforms the corresponding model trained on USPTO reactions only.