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

BMC Bioinformatics

Table 5 Performance of SMR-DDI and other featurizers on Task 3

Featurizer	AUPRC	AUROC	ACC	F1	Precision	Recall
MOL2VEC	0.296 ± 0.035	0.786 ± 0.031	0.343 ± 0.019	0.338 ± 0.017	0.352 ± 0.019	0.343 ± 0.019
gin_supervised_edgepred	0.233 ± 0.053	0.744 ± 0.029	0.345 ± 0.015	0.327 ± 0.015	0.343 ± 0.017	0.345 ± 0.015
ECFP	0.251 ± 0.023	0.767 ± 0.04	0.34 ± 0.014	0.323 ± 0.018	0.343 ± 0.01	0.34 ± 0.014
MACCKEYS	0.227 ± 0.07	0.71 ± 0.058	0.338 ± 0.023	0.317 ± 0.022	0.336 ± 0.019	0.338 ± 0.023
gin_supervised_infomax	0.228 ± 0.035	0.704 ± 0.038	0.339 ± 0.017	0.315 ± 0.019	0.339 ± 0.015	0.339 ± 0.017
gin_supervised_contextpred	0.244 ± 0.029	0.752 ± 0.028	0.344 ± 0.022	0.313 ± 0.016	0.336 ± 0.017	0.344 ± 0.022
gin_supervised_masking	0.225 ± 0.034	0.723 ± 0.03	0.328 ± 0.007	0.311 ± 0.008	0.339 ± 0.018	0.328 ± 0.007
ChemBERTa-77 M-MLM	0.239 ± 0.042	0.734 ± 0.029	0.334 ± 0.028	0.311 ± 0.026	0.336 ± 0.029	0.334 ± 0.028
ChemBERTa-77 M-MLR	0.249 ± 0.025	0.755 ± 0.027	0.318 ± 0.018	0.302 ± 0.016	0.319 ± 0.022	0.318 ± 0.018
SMR-DDI	0.154 ± 0.019	0.739 ± 0.041	0.305 ± 0.014	0.295 ± 0.013	0.312 ± 0.014	0.305 ± 0.014
ChemGPT-1.2B	0.165 ± 0.016	0.739 ± 0.035	0.279 ± 0.014	0.263 ± 0.018	0.271 ± 0.022	0.279 ± 0.014
ChemGPT-4 M	0.163 ± 0.019	0.749 ± 0.026	0.262 ± 0.021	0.254 ± 0.016	0.274 ± 0.015	0.262 ± 0.021

ISSN: 1471-2105