An ensemble machine learning model based on multiple filtering and supervised attribute clustering algorithm for...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Machine learning (ML) techniques excel at pattern detection in large datasets.
• ML is crucial in medical applications, particularly with genomic and proteomic data.
• Accurate cancer classification using bio-molecular data enhances diagnosis and treatment outcomes.

Purpose of the Study:

• To introduce a novel ensemble ML model for cancer classification.
• To address challenges of class imbalance and high dimensionality in microarray data.
• To improve diagnostic accuracy and identify significant biomarker genes in cancer samples.

Main Methods:

• Developed the Multiple Filtering and Supervised Attribute Clustering algorithm based Ensemble Classification (MFSAC-EC) model.
• Employed bootstrapping and oversampling to manage class imbalance.
• Utilized a supervised attribute clustering algorithm for feature selection, creating informative sub-datasets for ensemble classification via majority voting.

Main Results:

• The MFSAC-EC model demonstrated superior generalization performance and testing accuracy compared to existing models.
• The model successfully identified numerous important attributes and potential biomarker genes.
• Experimental validation was performed on high-dimensional microarray gene expression datasets for cancer classification.

Conclusions:

• The MFSAC-EC model offers a significant advancement in ML-based cancer classification.
• The approach effectively handles complex, imbalanced, and high-dimensional genomic datasets.
• The identified biomarker genes hold potential for future cancer diagnostics and targeted therapies.