Machine Learning Models for Cancer Research: A Narrative Review of Bulk RNA-Seq Applications.

Area of Science:

• Bioinformatics
• Computational Biology
• Oncology

Background:

• High-throughput sequencing data, particularly bulk RNA-seq, is crucial for identifying cancer-associated gene expression patterns.
• Machine learning (ML) algorithms efficiently process high-dimensional RNA-seq data to build diagnostic and prognostic models.
• The integration of ML and RNA-seq holds significant promise for advancing molecular medicine and biological discovery in oncology.

Purpose of the Study:

• To provide a comprehensive overview of bulk RNA-seq-based machine learning models in oncology.
• To offer practical guidance on workflow, algorithm selection, and study design for developing these models.
• To discuss bulk RNA-seq deconvolution as an alternative to single-cell RNA-seq for analyzing tumor cellularity.

Main Methods:

• Narrative review of existing literature on bulk RNA-seq and ML in oncology.
• Description of a complete workflow from data preprocessing to model validation.
• Discussion of deconvolution techniques for estimating cell type proportions from bulk RNA-seq data.

Main Results:

• Bulk RNA-seq combined with ML provides a cost-effective approach for cancer research.
• Practical recommendations for algorithm selection and study design are provided.
• Deconvolution methods offer a viable alternative to single-cell RNA-seq for assessing tumor composition.

Conclusions:

• The synergy between bulk RNA-seq and ML facilitates the development of reproducible diagnostic and prognostic models in oncology.
• These models have significant translational potential for improving patient outcomes.
• This review serves as a practical guide for researchers entering the field.