Double weighted k nearest neighbours for binary classification of high dimensional genomic data

Area of Science:

• Bioinformatics
• Computational Biology
• Machine Learning

Background:

• High-dimensional gene expression data presents challenges for classification due to numerous genes and limited samples.
• Existing methods struggle with prediction accuracy and computational efficiency on such datasets.

Purpose of the Study:

• To introduce a novel classification procedure, double weighted k nearest neighbours (DWKNN), specifically designed for high-dimensional gene expression data.
• To enhance classification accuracy and efficiency by focusing on informative genes.

Main Methods:

• DWKNN incorporates feature weights based on differential gene expression between classes.
• Weights are automatically adjusted to prioritize informative features and downplay non-informative ones.
• A two-fold weighted distance calculation strategy using an exponential function is employed for classification.

Main Results:

• Experimental evaluations demonstrate DWKNN's effectiveness in classifying gene expression datasets.
• DWKNN outperformed several established methods including standard kNN, weighted kNN, and Support Vector Machines (SVM).
• Performance was assessed using metrics like classification accuracy, Cohen's kappa, sensitivity, and F1-score.

Conclusions:

• DWKNN offers a robust and efficient approach for analyzing gene expression data.
• The method's ability to leverage informative genes leads to improved classification performance.
• DWKNN shows promise for advancing bioinformatics and computational biology research.