A DEEP LEARNING APPROACH FOR CANCER DETECTION AND RELEVANT GENE IDENTIFICATION

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Cancer detection from gene expression data is challenging due to high dimensionality and complexity.
• Clinical diagnosis and tumor-specific marker identification remain areas of uncertainty.
• Advanced computational methods are needed to analyze complex genomic datasets for cancer research.

Purpose of the Study:

• To develop a deep learning approach for enhanced cancer detection using gene expression profiles.
• To identify critical genes for accurate breast cancer diagnosis.
• To explore the utility of extracted functional features for cancer classification.

Main Methods:

• Utilized Stacked Denoising Autoencoder (SDAE) for deep feature extraction from high-dimensional gene expression data.
• Employed supervised classification models to evaluate the performance of extracted features in cancer detection.
• Analyzed SDAE connectivity matrices to identify highly interactive genes.

Main Results:

• Deep learning effectively extracted functional features from complex gene expression data.
• The extracted features demonstrated usefulness in supervised cancer detection models.
• A set of highly interactive genes was identified as potential breast cancer biomarkers.

Conclusions:

• The proposed deep learning approach offers a promising method for cancer detection.
• Identified interactive genes may serve as valuable biomarkers for breast cancer diagnosis.
• Further research is warranted to validate these potential biomarkers in clinical settings.