Kernel Based Nonlinear Dimensionality Reduction and Classification for Genomic Microarray.

Area of Science:

• Bioinformatics
• Genomic Research
• Computational Biology

Background:

• Genomic microarrays generate high-dimensional data, posing challenges for feature extraction and dimensionality reduction.
• Effective analysis of gene expression data is crucial for advancements in bioinformatics and medicinal research.

Purpose of the Study:

• To propose a nonlinear dimensionality reduction kernel method based on Locally Linear Embedding (LLE).
• To enhance LLE by incorporating a fuzzy K-nearest neighbors (KNN) algorithm for data denoising.
• To apply Support Vector Machine (SVM) with a kernel method for classifying genomic microarray data.

Main Methods:

• Locally Linear Embedding (LLE) with a kernel-based nonlinear dimensionality reduction approach.
• Fuzzy K-nearest neighbors (KNN) algorithm for denoising datasets, replacing the classical KNN in LLE.
• Kernel Support Vector Machine (SVM) for the classification of genomic microarray data.

Main Results:

• Demonstrated application of the proposed methods on two published DNA microarray datasets.
• The enhanced LLE method with fuzzy KNN showed improved performance in dimensionality reduction.
• Kernel SVM achieved high success rates in classifying the genomic microarray data.

Conclusions:

• The presented kernel-based LLE with fuzzy KNN offers a superior approach for analyzing high-dimensional genomic microarray data.
• The combined method effectively addresses challenges in feature extraction and dimensionality reduction.
• This approach shows high success rates and potential for advancing bioinformatics and medicinal research.