Fig. 3: Quality evaluations of the synthetic data. | Nature Communications

Fig. 3: Quality evaluations of the synthetic data.

From: RSGPT: a generative transformer model for retrosynthesis planning pre-trained on ten billion datapoints

Fig. 3: Quality evaluations of the synthetic data.

a Reaction products were randomly selected to generate the Tree map (TMAP). Blue and red dots represent products from synthetic data and USPTO-50k51,60, respectively. b Reactants were randomly selected to generate the Tree map (TMAP). Blue and red dots represent products from synthetic data and USPTO-50k, respectively. c Distribution of k based on 1500 randomly selected synthetic data. Here, k represents the number of templates that can match a given set of reactants R. d Blind evaluation of the validity of 100 USPTO-50k data entries and 100 synthetic data entries by experts.

Back to article page