GeneViT: Gene Vision Transformer with Improved DeepInsight for cancer classification

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Gene expression data analysis is vital for disease diagnosis and prognosis.
• High redundancy and noise in gene expression data pose challenges for accurate disease information extraction.
• Conventional machine learning and deep learning models have limitations in analyzing complex gene expression patterns.

Purpose of the Study:

• To introduce a novel method for classifying cancerous gene expression using a vision transformer architecture.
• To address the limitations of existing models in handling high-dimensional and noisy gene expression data.
• To explore the potential of vision transformers in the field of gene expression analysis.

Main Methods:

• Gene expression data is preprocessed using a stacked autoencoder for dimensionality reduction.
• The Improved DeepInsight algorithm transforms the reduced data into an image format.
• A vision transformer model is employed for the classification of cancerous gene expression data.

Main Results:

• The proposed vision transformer-based model demonstrates superior performance in classifying cancerous gene expression across ten benchmark datasets.
• Experimental results show that the model outperforms nine existing classification methods.
• t-SNE plots confirm the model's capability for distinctive feature learning.

Conclusions:

• The developed vision transformer method offers a powerful and effective approach for cancerous gene expression classification.
• This study highlights the potential of vision transformers in advancing gene expression data analysis for disease insights.
• The proposed method provides a promising tool for improved disease prognosis and diagnosis.