Table 2.

Feature comparison—performance of SVM over all versus common ADRs

Feature set	ADR_All				ADR_50+
	AUC	ACC	Precision	Recall	AUC	ACC	Precision	Recall
Chem	0.9054	0.9538	0.4337	0.4925	0.7659	0.8268	0.4539	0.5569
Bio	0.9069	0.9543	0.4324	0.5043	0.7729	0.8287	0.4666	0.5521
Pheno	0.9542	0.9678	0.6607	0.6460	0.9175	0.8891	0.6933	0.7142
Chem+bio	0.9098	0.9551	0.4623	0.5008	0.7849	0.8327	0.4776	0.5728
Chem+pheno	0.9526	0.9669	0.6488	0.6443	0.9141	0.8857	0.6757	0.7215
Chem+bio+pheno	0.9524	0.9669	0.6617	0.6306	0.9138	0.8856	0.6750	0.7227

ADR_All considers all ADRs and ADR_50+ are the common ADRs caused by at least 50 drugs. All AUC, ACC, Precision, and Recall are micro-averages across ADRs in the corresponding dataset.

ACC, accuracy; ADR, adverse drug reaction; AUC, area under the receiver operating characteristic curve; Bio, biological property; Chem, chemical structure; Pheno, phenotypic property; SVM, support vector machine.