# Machine Learning Methods for Cancer Classification Using Gene Expression Data: A Review

**Authors:** Fadi Alharbi, Aleksandar Vakanski  
**arXiv:** [2301.12222](https://arxiv.org/abs/2301.12222)  
**Date:** 2023-01-28  
**Summary page:** https://www.summarizepaper.com/en/paper/2301.12222v1/

## Key points

- Cancer is a complex group of diseases characterized by abnormal cell growth that can spread to different parts of the body, making it the second leading cause of death globally.
- Gene expression analysis is crucial for early cancer detection by providing insights into the biochemical processes and genetic characteristics of cells and tissues.
- DNA microarrays and RNA-sequencing methods are essential tools for quantifying gene expression levels and generating valuable data for computational analysis.
- Recent advancements in gene expression analysis for cancer classification involve machine learning techniques, with a focus on deep learning models due to their ability to identify unique gene patterns associated with various types of cancers.
- Popular deep neural network architectures such as multi-layer perceptrons, convolutional networks, recurrent networks, graph networks, and transformer networks are utilized in this study.
- Data collection methods for gene expression analysis and key datasets commonly used for supervised machine learning in this domain are discussed.
- Important techniques for feature engineering and data preprocessing to address the high dimensionality of gene expression data are highlighted.
- The study provides novel insights on graph neural networks (GNN) and transformer neural networks (TNN) for gene expression analysis not covered in prior works.
- The review contributes valuable insights into both conventional machine learning methods and recent deep learning approaches for gene expression analysis in cancer classification.

## Summary

Cancer is a complex group of diseases characterized by abnormal cell growth that can spread to different parts of the body, making it the second leading cause of death globally. Gene expression analysis plays a crucial role in early cancer detection by providing insights into the biochemical processes and genetic characteristics of cells and tissues. Deoxyribonucleic Acid (DNA) microarrays and Ribonucleic Acid (RNA)-sequencing methods are essential tools for quantifying gene expression levels and generating valuable data for computational analysis. This study explores recent advancements in gene expression analysis for cancer classification using machine learning techniques. Both traditional and deep learning-based approaches are reviewed, with a focus on the application of deep learning models due to their ability to identify unique gene patterns associated with various types of cancers. The survey covers relevant works utilizing popular deep neural network architectures such as multi-layer perceptrons, convolutional networks, recurrent networks, graph networks, and transformer networks. Additionally, the paper provides an overview of data collection methods for gene expression analysis and highlights key datasets commonly used for supervised machine learning in this domain. It also discusses important techniques for feature engineering and data preprocessing to address the high dimensionality of gene expression data resulting from a large number of genes present in samples. Furthermore, the authors list previous review papers related to gene expression analysis, offering comparative information on conventional machine learning approaches, feature engineering techniques, deep learning approaches like recurrent neural networks (RNN) and convolutional neural networks (CNN), as well as the use of microarray versus RNA-Seq data. The study distinguishes itself by providing novel insights not covered in prior works, including a comprehensive discussion on graph neural networks (GNN) and transformer neural networks (TNN) for gene expression analysis. Overall, this comprehensive review contributes valuable insights into both conventional machine learning methods and recent deep learning approaches for gene expression analysis in cancer classification. It also sheds light on modeling RNA-Seq data formats which have become dominant in recent years while discussing related feature engineering techniques not extensively covered in existing literature. The paper concludes with suggestions for future research directions in machine learning-based gene expression analysis for cancer classification.