NNBGWO-BRCA marker: Neural Network and binary grey wolf optimization based Breast cancer biomarker discovery...

Area of Science:

• Biomedical Informatics
• Computational Biology
• Genomics

Background:

• Breast cancer poses a significant mortality risk, necessitating early detection and precise classification.
• Multi-omics data integration is crucial for identifying breast cancer biomarkers and subtypes.
• Deep learning offers potential for analyzing complex multi-omics data but lacks robust feature selection methods.

Purpose of the Study:

• To propose a novel deep learning framework, NNBGWO-BRCAMarker, for discovering breast cancer biomarkers from multi-omics data.
• To leverage neural networks and grey wolf optimization for effective feature selection in biomarker discovery.
• To enable precise classification of breast cancer subtypes for targeted therapeutic strategies.

Main Methods:

• A two-phase framework: 1) Gene selection using feedforward neural network weights. 2) Further gene screening via binary grey wolf optimization.
• Utilized multi-omics data for comprehensive biomarker identification.
• Employed Support Vector Machine (SVM) with RBF kernel for classification accuracy assessment.

Main Results:

• Identified 80 potential breast cancer biomarkers using the NNBGWO-BRCAMarker framework.
• Achieved a classification accuracy of 0.9242 ± 0.03 for breast cancer subtypes using the identified biomarkers.
• Uncovered 25 druggable genes, 16 enriched pathways, and 8 prognostic genes through subsequent analyses.

Conclusions:

• The NNBGWO-BRCAMarker framework successfully identified key biomarkers for accurate breast cancer subtype classification.
• The discovered biomarkers and associated analyses provide valuable insights for clinical research and personalized medicine.
• This study highlights the potential of integrating deep learning with optimization algorithms for multi-omics-based biomarker discovery.