Machine learning-based plasma-derived extracellular vesicle signatures for digestive system cancers prediction

Area of Science:

• Oncology
• Molecular Diagnostics
• Bioinformatics

Background:

• Digestive system cancers (DSCs) are difficult to diagnose early, with limited accuracy from traditional biomarkers.
• Plasma-derived extracellular vesicle (PDEV) RNA is a promising diagnostic tool, but its use in DSCs requires further investigation.

Purpose of the Study:

• To develop and validate a machine learning-based diagnostic model for DSCs using PDEV RNA.
• To identify novel exosomal RNA biomarkers for non-invasive DSCs detection.

Main Methods:

• Integrated PDEV sequencing data from exoRBase 2.0 for 444 participants (326 DSCs patients, 118 healthy).
• Constructed and validated a PDEV-diagnostic model using machine learning algorithms (XGBoost) with 5-fold cross-validation.
• Assessed exosomal RNA features using bulk and single RNA-seq data.

Main Results:

• The XGBoost model achieved high diagnostic accuracy, with AUCs of 0.83 (training) and 0.94 (test sets).
• Identified nine exosomal RNA predictors (BANK1, MALAT1, FGA, UBR4, ILR-7, FGB, PLPP5, PCAT19, CIITA) for DSCs.
• Demonstrated the model's performance across training and test datasets.

Conclusions:

• Machine learning models utilizing PDEV RNA show high accuracy for DSCs diagnosis.
• The identified nine exosomal mRNAs/lncRNAs are promising non-invasive biomarkers for early DSCs detection.