Table 2.

Effect of the fingerprint parameters on model predictive performance. The row in bold corresponds to the optimal fingerprint characteristics identified in the outer grid search. The SI provides detailed information for every cross-validation iteration (S6_outer_inner_grid_details_rf.xlsx and S7_outer_inner_grid_details_rf.xlsx).

Fingerprint parameters		kNN		RF
Radius	Length	Optimal model parameters (n_neighbors, weights)	Cross- validation F1 (macro) score	Optimal model parameters (max_features, min_samples_split, n_estimators)	Cross- validation F1 (macro) score
2	1536	5, distance	0.771	0.01, 4, 200	0.850
2	2048	4, distance	0.781	0.002, 4, 250	0.845
2	2560	3, distance	0.784	0.01, 3, 150	0.853
3	1536	4, distance	0.766	0.02, 3, 150	0.838
3	2048	4, distance	0.781	0.005, 4, 200	0.834
3	2560	4, distance	0.780	0.02, 3, 250	0.842
4	1536	3, distance	0.771	0.005, 4, 250	0.825
4	2048	4, distance	0.779	0.2, 3, 250	0.830
4	2560	4, distance	0.778	0.02, 3, 100	0.839
5	1536	4, distance	0.773	0.02, 2, 150	0.823
5	2048	4, distance	0.776	0.15, 2, 200	0.821
5	2560	4, distance	0.776	0.1, 2, 300	0.831