DGDFS: Dependence Guided Discriminative Feature Selection for Predicting Adverse Drug-Drug Interaction

Adverse drug-drug interaction (ADDI) is referred to as a situation where the unpleasant or adverse effects caused by the co-administration of two drugs, which becomes a significant problem for public health. With the increasing availability of healthcare data, many methods are proposed for ADDI prediction. However, these methods usually work in a “nondiscriminatory” manner, i.e., they treat each feature without discrimination and equally incorporate all features into the predictive models. In practice, only a few features are essentially discriminative and relevant to ADDIs. In this paper, we propose a Dependence Guided Discriminative Feature Selection (DGDFS) model for ADDI prediction. In DGDFS, two drug attributes, molecular structure and side effect are adopted to model the adverse interaction among drugs and $l_{2,0}$-norm equality constraints are introduced to select discriminative molecular substructures and side effects for ADDI prediction. Besides, three dependence guided terms, i.e., the dependence between molecular structure and ADDI, the dependence between side effect and ADDI, and the dependence between molecular structure and side effect, are designed to guide feature selection. An iterative algorithm based on the alternating direction method of multipliers is developed for optimization. Experimental results indicate the effectiveness of DGDFS compared with fourteen baselines and its three variants.

1 Introduction

During pharmacotherapy, adverse drug reaction (ADR) is known as the harmful effects caused by individual drugs or drug combinations, which becomes an important clinical problem for public health and a significant financial burden for healthcare systems [1]. In the United States, ADRs are estimated to result in more than 2 million hospitalizations and 100,000 deaths per year and wipe about $136 billion from healthcare annual budget [2]. As one of the main concerns in ADRs, adverse drug-drug interaction is responsible for 30 percent of the reported ADRs and referred to as a situation where the unpleasant or adverse effects caused by the co-administration of two drugs [3], leading to dangerous reactions and inevitable damages to patients’ health [4]. A typical example is that Aspirin, an analgesic-antipyretic, taken along with Warfarin, a kind of blood-thinning drugs, may result in epigastric pain and gastrointestinal hemorrhage [5]. Although the detection of ADDIs is conducted in the vitro experiments during the preclinical stage of drug developments, there are still a number of new ADDIs discovered by accident after drugs hit the market [6]. Their occurrences primarily lie on the complex interactions between drugs and human body [7], which makes it difficult for the vitro experiments to simulate the vivo environments for exploring ADDIs in advance [8], [9]. These undetected ADDIs delay the disease treatments, aggravate the illness conditions, and increase the morbidity and mortality of patients, which brings an urgency to predict ADDIs before the medications are administered for preventing this life-threatening issue and enhancing the safety of pharmacotherapy in clinical treatments [2], [6].