Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Boosting drug-disease association prediction for drug repositioning via dual-feature extraction and cross-dual-domain decoding

Enqiang Zhu, Xiang Li, Chanjuan Liu, Nikhil R. Pal

TL;DR

This work addresses the challenge of accurately predicting drug–disease associations for repositioning by introducing DFDRNN, a dual-feature neural network that jointly encodes similarity and association information. The encoder uses two modules, SDDFE and CDDFE, with a self-attention mechanism to capture intra- and inter-domain relationships, while a cross-dual-domain decoder yields predictions in both drug and disease spaces. DFDRNN consistently outperforms six baselines on four datasets (AUROC up to $0.946$, AUPR up to $0.597$) and demonstrates strong generalization in leave-one-disease-out and case studies for Alzheimer's and Parkinson's diseases, supported by corroborating data sources. The method promises practical impact for rapid drug repurposing by leveraging dual-feature encoding and cross-domain decoding to reduce noise and improve the reliability of predicted associations.

Abstract

The extraction of biomedical data has significant academic and practical value in contemporary biomedical sciences. In recent years, drug repositioning, a cost-effective strategy for drug development by discovering new indications for approved drugs, has gained increasing attention. However, many existing drug repositioning methods focus on mining information from adjacent nodes in biomedical networks without considering the potential inter-relationships between the feature spaces of drugs and diseases. This can lead to inaccurate encoding, resulting in biased mined drug-disease association information. To address this limitation, we propose a new model called Dual-Feature Drug Repurposing Neural Network (DFDRNN). DFDRNN allows the mining of two features (similarity and association) from the drug-disease biomedical networks to encode drugs and diseases. A self-attention mechanism is utilized to extract neighbor feature information. It incorporates two dual-feature extraction modules: the single-domain dual-feature extraction (SDDFE) module for extracting features within a single domain (drugs or diseases) and the cross-domain dual-feature extraction (CDDFE) module for extracting features across domains. By utilizing these modules, we ensure more appropriate encoding of drugs and diseases. A cross-dual-domain decoder is also designed to predict drug-disease associations in both domains. Our proposed DFDRNN model outperforms six state-of-the-art methods on four benchmark datasets, achieving an average AUROC of 0.946 and an average AUPR of 0.597. Case studies on two diseases show that the proposed DFDRNN model can be applied in real-world scenarios, demonstrating its significant potential in drug repositioning.

Boosting drug-disease association prediction for drug repositioning via dual-feature extraction and cross-dual-domain decoding

TL;DR

This work addresses the challenge of accurately predicting drug–disease associations for repositioning by introducing DFDRNN, a dual-feature neural network that jointly encodes similarity and association information. The encoder uses two modules, SDDFE and CDDFE, with a self-attention mechanism to capture intra- and inter-domain relationships, while a cross-dual-domain decoder yields predictions in both drug and disease spaces. DFDRNN consistently outperforms six baselines on four datasets (AUROC up to , AUPR up to ) and demonstrates strong generalization in leave-one-disease-out and case studies for Alzheimer's and Parkinson's diseases, supported by corroborating data sources. The method promises practical impact for rapid drug repurposing by leveraging dual-feature encoding and cross-domain decoding to reduce noise and improve the reliability of predicted associations.

Abstract

The extraction of biomedical data has significant academic and practical value in contemporary biomedical sciences. In recent years, drug repositioning, a cost-effective strategy for drug development by discovering new indications for approved drugs, has gained increasing attention. However, many existing drug repositioning methods focus on mining information from adjacent nodes in biomedical networks without considering the potential inter-relationships between the feature spaces of drugs and diseases. This can lead to inaccurate encoding, resulting in biased mined drug-disease association information. To address this limitation, we propose a new model called Dual-Feature Drug Repurposing Neural Network (DFDRNN). DFDRNN allows the mining of two features (similarity and association) from the drug-disease biomedical networks to encode drugs and diseases. A self-attention mechanism is utilized to extract neighbor feature information. It incorporates two dual-feature extraction modules: the single-domain dual-feature extraction (SDDFE) module for extracting features within a single domain (drugs or diseases) and the cross-domain dual-feature extraction (CDDFE) module for extracting features across domains. By utilizing these modules, we ensure more appropriate encoding of drugs and diseases. A cross-dual-domain decoder is also designed to predict drug-disease associations in both domains. Our proposed DFDRNN model outperforms six state-of-the-art methods on four benchmark datasets, achieving an average AUROC of 0.946 and an average AUPR of 0.597. Case studies on two diseases show that the proposed DFDRNN model can be applied in real-world scenarios, demonstrating its significant potential in drug repositioning.
Paper Structure (22 sections, 20 equations, 5 figures, 6 tables)

This paper contains 22 sections, 20 equations, 5 figures, 6 tables.

Table of Contents

  1. Introduction
  2. Materials and Methods
  3. Datasets
  4. Model structure
  5. Three key networks
  6. Self-attention mechanism
  7. Encoder
  8. Decoder
  9. Optimization
  10. Experiments and Results
  11. Baseline methods
  12. Experimental parameter configuration
  13. Evaluation metrics
  14. DFDRNN's performance in cross-validation
  15. Effect of the number of nearest neighbors
  16. ...and 7 more sections

Figures (5)

  • Figure 1: DFDRNN generates the similarity feature and the association feature for drugs and diseases. The encoding process employs a multi-layer connected network architecture, where each layer performs SDDFE and CDDFE, while tracking the dynamic changes of dual-feature. A layer attention mechanism computes final features. The decoder then calculates drug-disease association scores separately for each domain.
  • Figure 2: (a)Line plot of AUROC as a function of top-$t$ on four benchmark datasets; (b)Line plot of AUPR as a function of top-$t$ on four benchmark datasets; (c)Line plot of the average AUROC as a function of top-$t$ on four datasets; (d)Line plot of the average AUPR as a function of top-$t$ on four datasets.
  • Figure 3: non-cross dual-domain decoder
  • Figure 4: Performance of different decoding methods: (a) AUROC heatmap; (b) AUPR heatmap.
  • Figure 5: The performance of all approaches in predicting candidate drugs for new diseases on Gdataset: (a) ROC curves illustrating the prediction outcomes of DFDRNN alongside other competitive models; (b) PR curves illustrating the prediction outcomes of DFDRNN alongside other competitive models.