DDIT: An Online Predictor for Multiple Clinical Phenotypic Drug-Disease Associations

Abstract

Background: Drug repurposing provides an effective method for high-speed, low-risk drug development. Clinical phenotype-based screening exceeded target-based approaches in discovering first-in-class small-molecule drugs. However, most of these approaches predict only binary phenotypic associations between drugs and diseases; the types of drug and diseases have not been well exploited. Principally, the clinical phenotypes of a known drug can be divided into indications (Is), side effects (SEs), and contraindications (CIs). Incorporating these different clinical phenotypes of drug-disease associations (DDAs) can improve the prediction accuracy of the DDAs. Methods: We develop Drug Disease Interaction Type (DDIT), a user-friendly online predictor that supports drug repositioning by submitting known Is, SEs, and CIs for a target drug of interest. The dataset for Is, SEs, and CIs was extracted from PREDICT, SIDER, and MED-RT, respectively. To unify the names of the drugs and diseases, we mapped their names to the Unified Medical Language System (UMLS) ontology using Rest API. We then integrated multiple clinical phenotypes into a conditional restricted Boltzmann machine (RBM) enabling the identification of different phenotypes of drug-disease associations, including the prediction of as yet unknown DDAs in the input. Results: By 10-fold cross-validation, we demonstrate that DDIT can effectively capture the latent features of the drug-disease association network and represents over 0.217 and over 0.072 improvement in AUC and AUPR, respectively, for predicting the clinical phenotypes of DDAs compared with the classic K-nearest neighbors method (KNN, including drug-based KNN and disease-based KNN), Random Forest, and XGBoost. By conducting leave-one-drug-class-out cross-validation, the AUC and AUPR of DDIT demonstrated an improvement of 0.135 in AUC and 0.075 in AUPR compared to any of the other four methods. Within the top 10 predicted indications, side effects, and contraindications, 7/10, 9/10, and 9/10 hit known drug-disease associations. Overall, DDIT is a useful tool for predicting multiple clinical phenotypic types of drug-disease associations.

Keywords: contraindication; drug repositioning; indication; machine learning; phenotypic types of drug-disease associations; restricted Boltzmann machine; side effect.

FIGURE 1

Flowchart of DDIT modeling. First, we collected different phenotypic types from PREDICT, SIDER, and MED-RT, respectively, followed by mapping the disease and drug names to UML ontologies. We then constructed a multidimensional network with nodes presenting drug or disease, and with the edge representing indication, side effect, and contraindication. The formatted data were input into a restricted Boltzmann machine (RBM) and the output reconstructed the input. We here take five drugs and four diseases as example. The left matrices represent three types of DDAs. For each matrix, the row represents drugs, while the column represents disease. The square is white if A_ij = 1, which means the drug and disease have indication/side effect/contraindication, black if otherwise. The right matrix is the output probability. The y-axis represents drug, the x-axis represents disease, and the z-axis represents the three types of DDAs. As the probability is in the range from 0 to 1, we color them in gray. I, indications; SE, side effects; CI, contraindications.

FIGURE 2

An RBM model. n and m are the number of visible units and hidden units, respectively. a is the bias of visible variables, while b is the bias of hidden variables. W is the connection weight matrix between each visible unit and each hidden unit.

FIGURE 3

A toy example of constructing conditional RBMs for two drugs and four diseases. The RBMs for both drug 1 and drug 2 share the same parameters.

FIGURE 4

ROC and PR curve of DDIT compared with permutation test and single phenotypic type data only. (A,D) ROC and PR curves comparing DDIT with permutation tests and indication data only for indication predictions of known drugs. (B,E) ROC and PR curves comparing DDIT with permutation test and side effect data only for side effect predictions of known drugs. (C,F) ROC and PR curves comparing DDIT with permutation test and contraindication data only for contraindication predictions of known drugs.

FIGURE 5

ROC and PR curves of DDIT compared with drug-based KNN, disease-based KNN, Random Forest classifier, and XGBoost. (A,D) ROC and PR curves comparing DDIT with drug-based KNN, disease-based KNN, Random Forest, and XGBoost for novel indication prediction of known drugs. (B,E) ROC and PR curves comparing DDIT with drug-based KNN, disease-based KNN, Random Forest, and XGBoost for novel side effect prediction of known drugs. (C,F) ROC and PR curves comparing DDIT with drug-based KNN, disease-based KNN, Random Forest, and XGBoost for novel contraindication prediction of known drugs.

FIGURE 6

DDIT web interface. (A) Home page. (B) Prediction page. Predicting clinical phenotypic types based on drug profiles submitted by users. (C) Search page. Left: search drug’s Is, SEs, and CIs by drug’s name; right: search for drugs that can cure, cause the disease, or as contraindicated in people with the disease by disease’s name. (D) eDoctor page. Provide recommended drugs for patients with underlying diseases.