Assessment of MicroRNAs Associated with Tumor Purity by Random Forest Regression

Area of Science:

• Oncology
• Genomics
• Bioinformatics

Background:

• Tumor purity, the proportion of tumor cells in samples, is crucial for understanding the tumor microenvironment.
• Predicting tumor purity is vital, but methods using microRNAs (miRNAs) are underdeveloped.

Purpose of the Study:

• To predict tumor purity using miRNA expression data across 16 TCGA tumor types.
• To identify key miRNAs for accurate tumor purity prediction and explore their biological relevance.

Main Methods:

• Utilized random forest regression to analyze miRNA expression data from The Cancer Genome Atlas (TCGA).
• Identified high-feature-importance miRNAs and assessed predictive performance using subsets of these miRNAs.
• Investigated genes targeted by predictive miRNAs to understand their association with cancer pathways.

Main Results:

• miRNA expression data effectively predicted tumor purity across various cancer types.
• A subset of 10 highly informative miRNAs achieved predictive performance comparable to using all miRNAs.
• Target genes of these miRNAs were significantly enriched in immune and cancer-related pathways.

Conclusions:

• miRNA expression profiles are valuable for predicting tumor purity.
• A small panel of 10 miRNAs shows potential as biomarkers for tumor purity assessment.
• These findings enhance understanding of the tumor microenvironment and miRNA roles in cancer.