Deep learning approach for cancer subtype classification using high-dimensional gene expression data.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer subtype classification is crucial for research and personalized treatment.
• Current methods using gene expression data struggle with small sample sizes and high-dimensional, sparse data.
• Effective classification requires methods that can handle data limitations.

Purpose of the Study:

• To develop a deep learning approach for accurate cancer subtype classification.
• To address the challenges of small sample sizes and high-dimensional, sparse gene expression data.
• To improve the performance of cancer classification beyond existing methods.

Main Methods:

• A novel deep learning approach, DCGN, combining Convolutional Neural Network (CNN) and Bidirectional Gated Recurrent Unit (BiGRU).
• Utilized Synthetic Minority Oversampling Technique (SMOTE) to equalize imbalanced datasets.
• CNN extracts local features from high-dimensional data; BiGRU analyzes deep features and retains information.

Main Results:

• DCGN demonstrated superior performance compared to seven other cancer subtype classification methods.
• Achieved more satisfactory classification results on breast and bladder cancer gene expression datasets.
• The combined CNN and BiGRU architecture effectively overcame data sparsity and small sample size limitations.

Conclusions:

• The proposed DCGN deep learning model offers a robust solution for cancer subtype classification.
• DCGN effectively handles high-dimensional, sparse gene expression data from limited samples.
• This approach holds promise for advancing cancer research and personalized medicine.