Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

CADGL: Context-Aware Deep Graph Learning for Predicting Drug-Drug Interactions

Azmine Toushik Wasi, Taki Hasan Rafi, Raima Islam, Serbetar Karlo, Dong-Kyu Chae

TL;DR

The paper tackles the challenge of accurate and generalizable DDI prediction by proposing CADGL, a context-aware deep graph learning framework built on a customized variational graph autoencoder. CADGL uses a two-encoder architecture (context-aware graph encoder and latent information encoder) plus an MLP decoder, augmented with Local Context and Molecular Context preprocessors and a self-supervised graph attention module, to extract robust structural and physico-chemical features. It achieves state-of-the-art performance on the DrugBankWishart2018-ae dataset and demonstrates the ability to predict clinically valuable novel DDIs, supported by ablation analyses and case studies. The work promises to accelerate early-stage drug discovery by providing reliable, interpretable links between drugs while acknowledging limitations and the need for further clinical validation.

Abstract

Examining Drug-Drug Interactions (DDIs) is a pivotal element in the process of drug development. DDIs occur when one drug's properties are affected by the inclusion of other drugs. Detecting favorable DDIs has the potential to pave the way for creating and advancing innovative medications applicable in practical settings. However, existing DDI prediction models continue to face challenges related to generalization in extreme cases, robust feature extraction, and real-life application possibilities. We aim to address these challenges by leveraging the effectiveness of context-aware deep graph learning by introducing a novel framework named CADGL. Based on a customized variational graph autoencoder (VGAE), we capture critical structural and physio-chemical information using two context preprocessors for feature extraction from two different perspectives: local neighborhood and molecular context, in a heterogeneous graphical structure. Our customized VGAE consists of a graph encoder, a latent information encoder, and an MLP decoder. CADGL surpasses other state-of-the-art DDI prediction models, excelling in predicting clinically valuable novel DDIs, supported by rigorous case studies.

CADGL: Context-Aware Deep Graph Learning for Predicting Drug-Drug Interactions

TL;DR

The paper tackles the challenge of accurate and generalizable DDI prediction by proposing CADGL, a context-aware deep graph learning framework built on a customized variational graph autoencoder. CADGL uses a two-encoder architecture (context-aware graph encoder and latent information encoder) plus an MLP decoder, augmented with Local Context and Molecular Context preprocessors and a self-supervised graph attention module, to extract robust structural and physico-chemical features. It achieves state-of-the-art performance on the DrugBankWishart2018-ae dataset and demonstrates the ability to predict clinically valuable novel DDIs, supported by ablation analyses and case studies. The work promises to accelerate early-stage drug discovery by providing reliable, interpretable links between drugs while acknowledging limitations and the need for further clinical validation.

Abstract

Examining Drug-Drug Interactions (DDIs) is a pivotal element in the process of drug development. DDIs occur when one drug's properties are affected by the inclusion of other drugs. Detecting favorable DDIs has the potential to pave the way for creating and advancing innovative medications applicable in practical settings. However, existing DDI prediction models continue to face challenges related to generalization in extreme cases, robust feature extraction, and real-life application possibilities. We aim to address these challenges by leveraging the effectiveness of context-aware deep graph learning by introducing a novel framework named CADGL. Based on a customized variational graph autoencoder (VGAE), we capture critical structural and physio-chemical information using two context preprocessors for feature extraction from two different perspectives: local neighborhood and molecular context, in a heterogeneous graphical structure. Our customized VGAE consists of a graph encoder, a latent information encoder, and an MLP decoder. CADGL surpasses other state-of-the-art DDI prediction models, excelling in predicting clinically valuable novel DDIs, supported by rigorous case studies.
Paper Structure (28 sections, 14 equations, 4 figures, 4 tables, 1 algorithm)

This paper contains 28 sections, 14 equations, 4 figures, 4 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Works
  3. Drug-Drug Interactions
  4. Context-Aware Learning in Drug-Interaction
  5. Autoencoders in Drug-Interaction
  6. Method and Algorithm
  7. Problem Formulation
  8. Context-Aware Deep Graph Encoder
  9. Local context processor (LCP).
  10. Molecular context processor (MCP).
  11. Self-supervised graph attention network (SSGAttn).
  12. Graph normalization layer.
  13. Latent Information Encoder and MLP Decoder
  14. Training Objectives
  15. Cross-entropy loss (CE Loss)
  16. ...and 13 more sections

Figures (4)

  • Figure 1: Drug-Drug Interactions Prediction Scenario
  • Figure 2: The overall framework of our CADGL. After taking input, structural data is processed with local and molecular context processors. These are passed into self-supervised graph attention for graph encoding. The output passes through two FC blocks with properties data for developing latent distribution and generate encodings. An MLP decoder extracts information, while a classifier gauges potential link likelihoods.
  • Figure 3: Validation Loss, AUROC and AUPRC score visualization of context pre-processors for first 80 epochs.
  • Figure 4: Case study of new DDIs: (1) Ziprasidone and Ribociclib, (2) Secobarbital and Arotinolol, (3) Ajmaline and Ergocalciferol, (4) Terlipressin and Amitriptyline, (5) Treprostinil and Methylphenobarbital, and (3) Canagliflozin and Carvedilol.