Figure 1.
Overview of the main workflow in this paper.
First, data were integrated from multiple sources, including network data (drug-ADR associations) and intrinsic data (chemical structures and ATC taxonomies of drugs and MedDRA taxonomies of ADRs). Next, topological features and intrinsic features were constructed based on network data and intrinsic data, respectively, and then integrated features were constructed by integrating topological features with intrinsic features, Finally different algorithms were selected to construct models to predict ADR, and comparative analyses were performed for features, algorithms and prediction results based on modeling experiments.
Figure 2.
Degree distributions of drugs and ADRs.
The left panel depicts the histograms of the degrees of drugs. The right panel depicts the histograms of the degrees of ADRs.
Table 1.
Statistics for the drug-ADE networks.
Table 2.
AUC scores of the models built with different topological features.
Table 3.
AUPR scores of the models built with different topological features.
Figure 3.
Distribution of prediction scores for different types of drug-ADR pairs.
The histograms of distributions of prediction scores of models built by four algorithms are shown. In each sub panel, the blue, green, yellow and red histograms represent the distributions of prediction scores for low degree drug- low degree ADRs, high degree drug- low degree ADRs, low degree drug- high degree ADRs and high degree drug- high degree ADRs, respectively.
Table 4.
The performances of the optimal models validated by prospective evaluation.