Predicting miRNA-Drug Interactions Based on Multi-source Feature Fusion of Heterogeneous Network.

Area of Science:

• Biochemistry
• Bioinformatics
• Computational Biology

Background:

• Cancer treatment resistance is a major therapeutic hurdle.
• MicroRNA (miRNA) expression significantly influences drug sensitivity and resistance.
• Accurate prediction of miRNA-drug interactions (MDIs) is vital for understanding drug resistance mechanisms.

Purpose of the Study:

• To develop an innovative computational framework for predicting miRNA-drug interactions (MDIs).
• To enhance understanding of drug resistance mechanisms in cancer therapy.

Main Methods:

• Proposed MSFFMDI, a dual-channel multi-source feature fusion framework utilizing heterogeneous networks.
• Channel 1: Attribute feature extraction using k-mer/word2vec for miRNAs and Graph Isomorphism Network/mol2vec for drugs.
• Channel 2: Topological feature extraction via heterogeneous networks, Graph Attention Network, and multi-scale Convolutional Neural Network.

Main Results:

• MSFFMDI demonstrated excellent predictive performance on two independent datasets.
• Experimental results confirmed the robustness and effectiveness of the proposed method.
• Case studies further validated the framework's practical applicability.

Conclusions:

• MSFFMDI offers a powerful and interpretable approach for predicting potential miRNA-drug interactions.
• The framework provides valuable insights into the complex mechanisms underlying cancer drug resistance.
• This method can aid in the development of more effective cancer therapies.