Predicting Driver Genes From Multi-Omics Data Using Hierarchical Multi-Feature Synergy Model.

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Cancer development involves complex genetic factors, with abnormal cancer driver genes playing a key role.
• Identifying driver genes is crucial for understanding cancer pathology and developing targeted therapies.
• Existing methods for driver gene identification often lack accuracy due to ignoring gene synergism and feature importance.

Purpose of the Study:

• To propose a novel method, HMFS (Hierarchical Multi-Feature Synergy), for accurate cancer driver gene identification.
• To address limitations in current methods by incorporating gene synergism and feature importance.
• To enhance the understanding of cancer pathogenic mechanisms and facilitate targeted therapy development.

Main Methods:

• Constructing a hypergraph using Node2vec and K-means algorithms.
• Extracting the Mutation Aggregation Coefficient by analyzing topological features and gene mutual exclusion.
• Performing differential expression analysis using miRNA and mRNA data, followed by feature fusion using the Hierarchical Multi-Feature Synergy model.

Main Results:

• HMFS demonstrated superior performance across all evaluation metrics on three real cancer datasets.
• The proposed method significantly outperformed seven representative existing driver gene identification methods.
• The Hierarchical Multi-Feature Synergy model effectively integrates multiple features for improved identification accuracy.

Conclusions:

• HMFS offers a more accurate and robust approach to cancer driver gene identification.
• The method's ability to capture gene synergism and feature importance represents a significant advancement.
• This work provides a valuable tool for cancer research, aiding in the elucidation of disease mechanisms and the development of precision medicine strategies.