Application of committee kNN classifiers for gene expression profile classification.

Area of Science:

• Bioinformatics
• Machine Learning
• Genomics

Background:

• Microarray gene expression profiling is crucial for understanding disease mechanisms.
• Accurate classification of gene expression data is essential for diagnosing and treating diseases like leukaemia.
• Individual k-Nearest Neighbour (kNN) classifiers may have limitations in complex datasets.

Purpose of the Study:

• To develop and evaluate a novel two-class classification system for microarray gene expression data.
• To improve classification accuracy and stability compared to individual kNN classifiers.
• To apply the developed system to classify leukaemia patient gene expression profiles.

Main Methods:

• A committee of five k-Nearest Neighbour (kNN) classifiers with varying architectures was developed.
• Each kNN classifier within the committee utilized a distinct subset of features.
• The system incorporated a sequence of straightforward data preprocessing steps.
• The classification system was applied to microarray gene expression profiles from leukaemia patients.

Main Results:

• The committee-based kNN classification system demonstrated effective performance in classifying microarray gene expression data.
• The committee approach consistently outperformed individual kNN classifiers.
• Significant improvements in both classification accuracy and stability were observed with the committee method.

Conclusions:

• The developed committee of kNN classifiers offers a robust and effective method for gene expression data classification.
• This approach provides enhanced accuracy and stability, outperforming single kNN models.
• The system shows promise for applications in leukaemia patient data analysis and potentially other genomic studies.