Fig. 4
From: DeepSA: a deep-learning driven predictor of compound synthesis accessibility
Prediction results for 1,800 compounds in TS3. AÂ Score distribution histogram for the results from DeepSA over TS3. B The bubble plot of SMILES length versus synthetic accessibility prediction for 1800 compounds in TS3. The vertical coordinates represent the prediction outcome of the model. True_HS represents true positives, True_ES is true negatives, False_HS indicates false positives, and False_ES represents false negatives. Bubble size represents the number of observations in the SMILES length intervals. The depth of the bubble color represents the proportion of the prediction outcome in each bin of the SMILES length of compounds