Table 1.

Overview of the optimization done on the model, performance (R², RMSE, Spearman ρ) on the test set constructed for this challenge

Model	Description	R2	RMSE	Spearman ρ
–	AlogP	0.83 [0.71,0.90]	0.73 [0.55,0.93]	0.90 [0.85,0.94]
–	XlogP3	0.85 [0.75,0.92]	0.67 [0.48,0.87]	0.91 [0.87,0.95]
–	S+ logP	0.95 [0.91,0.97]	0.40 [0.32,0.48]	0.97 [0.94,0.98]
1	default	0.93 [0.89,0.96]	0.45 [0.36,0.57]	0.96 [0.94,0.97]
2	1 + rdkit	0.93 [0.89,0.96]	0.45 [0.37,0.55]	0.96 [0.94,0.98]
3	rdkit only	0.88 [0.82,0.92]	0.60 [0.50,0.70]	0.94 [0.91,0.96]
4	1 + ChEMBL merged	0.88 [0.81,0.92]	0.60 [0.51,0.71]	0.94 [0.92,0.96]
5	1 + ChEMBL separate	0.93 [0.88,0.95]	0.47 [0.38,0.58]	0.96 [0.94,0.98]
6	5 + AZ_logD7.4	0.94 [0.91,0.96]	0.42 [0.35,0.50]	0.97 [0.95,0.97]
7	5 + AZ_ADME	0.94 [0.90,0.96]	0.44 [0.36,0.51]	0.97 [0.95,0.98]
8	6 + hyperopt parameters	0.93 [0.88,0.95]	0.47 [0.39,0.58]	0.96 [0.94,0.97]
9	6 + S+ logP/logD7.4 as tasks	0.95 [0.93,0.97]	0.38 [0.32,0.44]	0.97 [0.96,0.98]
10	6 + S+ logP/logD7.4 as descriptors	0.95 [0.92,0.97]	0.39 [0.34,0.44]	0.97 [0.96,0.98]
11	1, ensemble of 10	0.94 [0.89,0.96]	0.44 [0.35,0.55]	0.96 [0.94,0.98]
12	9, ensemble of 10	0.95 [0.92,0.97]	0.39 [0.33,0.46]	0.97 [0.96,0.98]

The ordinal model numbers in the left-most column indicate the sequence in which the models were developed: for example model 6 (5 + AZ_logD7.4) means that the settings/data of model 5 were used and the AZ_logD7.4 data were added. The 95% confidence interval for the different performance metrics is shown between square brackets