Machine Learning Analysis of RNA-seq Data for Diagnostic and Prognostic Prediction of Colon Cancer

Area of Science:

• Bioinformatics and computational biology
• Genomics and transcriptomics
• Machine learning in healthcare

Background:

• Omics data integration with machine learning (ML) aids disease prediction.
• RNA sequencing (RNA-seq) is a gold standard for transcriptomics analysis.
• ML and deep learning (DL) are increasingly used for healthcare predictions.

Purpose of the Study:

• Develop ML and DL models for colon cancer prediction and classification.
• Analyze extracellular vesicle RNA-seq data from healthy and colon cancer patients.
• Compare the performance of canonical ML and DL classifiers.

Main Methods:

• Processed RNA-seq data from extracellular vesicles.
• Utilized canonical ML classifiers: kNN, LMT, RT, RC, RF.
• Employed DL models: 1-D CNN, LSTM, BiLSTM, with GA for hyper-parameter optimization.

Main Results:

• Canonical ML models (RC, LMT, RF) achieved 97.33% accuracy for cancer prediction.
• Random Forest (RF) reached 97.33% accuracy for cancer stage classification.
• Deep learning models showed high performance, with 1-D CNN at 97.67% for prediction and BiLSTM at 98% for stage classification.

Conclusions:

• Both canonical ML and DL models demonstrate significant potential in colon cancer prediction and classification.
• Model performance can vary based on the number of features utilized.
• RNA-seq data combined with advanced algorithms offers a promising avenue for clinical applications.