Learning self-supervised molecular representations for drug–drug interaction prediction

BMC Bioinformatics

Table 3 Performance of SMR-DDI and other featurizers on Task 1

Featurizer	AUPRC	AUROC	ACC	F1	Precision	Recall
ECFP	0.942 ± 0.02	0.996 ± 0.002	0.954 ± 0.002	0.954 ± 0.002	0.954 ± 0.002	0.954 ± 0.002
gin_supervised_infomax	0.91 ± 0.013	0.994 ± 0.002	0.93 ± 0.003	0.93 ± 0.003	0.93 ± 0.003	0.93 ± 0.003
gin_supervised_contextpred	0.902 ± 0.02	0.993 ± 0.003	0.928 ± 0.003	0.928 ± 0.003		0.928 ± 0.003	0.928 ± 0.003
ChemBERTa-77 M-MLM	0.918 ± 0.008	0.993 ± 0.004	0.924 ± 0.006	0.924 ± 0.006	0.925 ± 0.005	0.924 ± 0.006
gin_supervised_masking	0.9 ± 0.023	0.993 ± 0.003	0.923 ± 0.006	0.923 ± 0.006		0.923 ± 0.005	0.923 ± 0.006
gin_supervised_edgepred	0.918 ± 0.013	0.992 ± 0.003	0.923 ± 0.003	0.923 ± 0.003	0.924 ± 0.003	0.923 ± 0.003
ChemBERTa-77 M-MLR	0.917 ± 0.016	0.992 ± 0.004	0.907 ± 0.007	0.908 ± 0.007	0.908 ± 0.007	0.908 ± 0.007
MACCKEYS	0.919 ± 0.016	0.994 ± 0.003	0.892 ± 0.022	0.893 ± 0.022	0.895 ± 0.02	0.892 ± 0.022
SMR-DDI	0.9 ± 0.005	0.992 ± 0.003	0.877 ± 0.014	0.877 ± 0.014	0.88 ± 0.013	0.877 ± 0.014
MOL2VEC	0.91 ± 0.006	0.992 ± 0.002	0.869 ± 0.028	0.869 ± 0.027	0.873 ± 0.023	0.869 ± 0.028
ChemGPT-4	0.875 ± 0.017	0.993 ± 0.002	0.847 ± 0.026	0.848 ± 0.026	0.854 ± 0.022	0.847 ± 0.026
ChemGPT-1	0.877 ± 0.017	0.99 ± 0.004	0.839 ± 0.047	0.839 ± 0.047	0.846 ± 0.041	0.839 ± 0.047

ISSN: 1471-2105