Machine learning-based quantitative prediction of drug exposure in drug-drug interactions using drug label information

Abstract

Many machine learning techniques provide a simple prediction for drug-drug interactions (DDIs). However, a systematically constructed database with pharmacokinetic (PK) DDI information does not exist, nor is there a machine learning model that numerically predicts PK fold change (FC) with it. Therefore, we propose a PK DDI prediction (PK-DDIP) model for quantitative DDI prediction with high accuracy, while constructing a highly reliable PK-DDI database. Reliable information of 3,627 PK DDIs was constructed from 3,587 drugs using 38,711 Food and Drug Administration (FDA) drug labels. This PK-DDIP model predicted the FC of the area under the time-concentration curve (AUC) within ± 0.5959. The prediction proportions within 0.8-1.25-fold, 0.67-1.5-fold, and 0.5-2-fold of the AUC were 75.77, 86.68, and 94.76%, respectively. Two external validations confirmed good prediction performance for newly updated FDA labels and FC from patients'. This model enables potential DDI evaluation before clinical trials, which will save time and cost.

Fig. 1. Drug-drug interaction (DDI) Prediction Pipeline Overview.

(Step 1) Reliable Food and Drug Administration (FDA) drug labels were used through the DailyMed website to build the pharmacokinetic (PK)-DDI dataset. A total of 38,711 FDA drug labels were obtained (Evaluation date: May 2020) from sentences/pictures/tables in the clinical pharmacology and drug interaction sections. (Step 2) Information on various drug properties from DrugBank (Evaluation date: March 2021) was obtained. Drug properties data may be arranged around the perpetrator and the victim drugs, and various polypeptides are radially linked. Polypeptide-PD (pharmacodynamics)-Drug-Type (PPDT) tokenization was proposed to represent drug pairs. A bag-of-words containing 2830 unique tokens was obtained. Each drug-drug pair was encoded as a 2830-dimensional vector through normalization with a Term Frequency-Inverse Document Frequency (tf-idf) matrix of bag-of-words. (Step 3) The Bagged (Bootstrap Aggregation) Tree method was used as an application model. The tree consisted of 615 branches and had 308 nodes for which fold change values were determined. (Step 4) A standalone application PK-DDI prediction (PK-DDIP) model is provided. Through this application, users may obtain predicted and reported fold change values, drug polypeptide information and its plot, single-nucleotide polymorphisms action, and alternative drug recommendation information at the 4^th anatomical therapeutic chemical level.

Fig. 2. Model performance.

a Distribution of predicted and labeled drug-drug interactions (DDIs) according to the Food and Drug Administration’s (FDA) classification criteria. A strong DDI means that the perpetrator drug increases the area under the time-concentration curve (AUC) of the victim drug by more than 5-fold or decreases the AUC to less than 0.2-fold. In moderate DDI, the perpetrator increases the victim drug AUC by 2- to 5-fold or decreases the victim drug AUC by 0.2 to 0.5-fold. When weak DDI occurs, the perpetrator increases the victim drug AUC by 1.25- to 2-fold or decreases it by 0.5- to 0.8-fold. The AUC fold change (FC) between 0.8- and 1.25-fold, which does not belong to any criteria, is defined as a negative DDI. b Heatmap for the predicted percentage of DDI classes correctly called among each DDI class in the label. Cells with higher percentages are colored red for each prediction class. c Percentage rank of a given value in a data set. d Scatter chart. e Evaluation for quantitative AUC FC. Case 1-1: {0.8 × FC_lab ≤ FC_pre} ∧ {1.25 × FC_lab ≥ FC_pre}. Case 1-2: {0.67 × FC_lab ≤ FC_pre} ∧ {1.5 × FC_lab ≥ FC_pre}. Case 1-3: {0.5 × FC_lab ≤ FC_pre} ∧ {2 × FC_lab ≥ FC_pre}. Case 2: Class_lab = Class_pre. Case 3: Case 1-1 ∨ Case 2. Lab, Label from FDA; Pre, Prediction from DDI prediction.

Fig. 3. Comparison with real patients’ result.

The comparison of pharmacokinetic drug-drug interaction prediction (PKDDIP) model results (predicted fold change values) and observed real-world patients’ results using tacrolimus as a victim drug in a tertiary hospital clinical data warehouse. SNUH, Seoul National University Hospital.

Fig. 4. Standalone application.

ATC anatomical therapeutic chemical, AUC area under the time-concentration curve, PK-DDIP pharmacokinetic drug-drug interaction prediction.