Unsupervised Learning Framework With Multidimensional Scaling in Predicting Epithelial-Mesenchymal Transitions.

Area of Science:

• Genomics
• Computational Biology
• Cancer Research

Background:

• Clustering tumor metastasis samples from large gene expression datasets is challenging, especially with limited samples.
• Epithelial-mesenchymal transition (EMT) is a critical mechanism driving tumor metastasis.
• Existing clustering methods struggle with high-dimensional gene expression data and small sample sizes.

Purpose of the Study:

• To develop a novel computational model for predicting EMT.
• To accurately cluster tumor metastasis samples using gene expression data.
• To improve the understanding of tumor metastasis mechanisms through gene expression analysis.

Main Methods:

• Proposed a novel model integrating multidimensional scaling (MDS), entropy, and random matrix theory.
• Determined the optimal reduced dimension for low-dimensional space representation.
• Developed a feature extraction method for identifying significant genes.

Main Results:

• The proposed model demonstrated superior performance compared to state-of-the-art clustering methods.
• Successfully predicted EMT in breast cancer gene expression data.
• Identified significant genes for predicting tumor metastasis.

Conclusions:

• The novel model provides an effective approach for clustering gene expression data in cancer research.
• The method enhances the prediction of EMT and tumor metastasis.
• This work contributes to advancing computational strategies in cancer genomics.