Glioma stages prediction based on machine learning algorithm combined with protein-protein interaction networks

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Glioma, a lethal nervous system cancer, has unclear molecular mechanisms.
• Understanding glioma's occurrence and development is crucial.
• This study aims to elucidate glioma's molecular mechanisms.

Purpose of the Study:

• To reveal molecular mechanisms of glioma using protein-protein interaction (PPI) networks and machine learning.
• To identify key differentially expressed genes (DEGs) involved in glioma progression.
• To develop predictive models for glioma grading.

Main Methods:

• Screening and selection of key DEGs using PPI networks.
• Application of five machine learning methods for glioma stage prediction.
• Analysis of selected genes using PPI networks, Gene Ontology (GO), and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways.

Main Results:

• Identified 19 genes (grade I-II), 21 genes (grade II-III), and 20 genes (grade III-IV).
• Developed predictive models: Complement Naive Bayes (grade II-III, 72.8% accuracy), Random Forest (grade I-II, 97.1%; grade III-IV, 83.2%).
• Functional analysis via PPI, GO, and KEGG improved understanding of DEGs in glioma growth.

Conclusions:

• Machine learning combined with PPI networks, GO, and KEGG analyses enhances understanding of glioma growth mechanisms.
• Identified key genes may guide glioma research and understanding.
• The study provides valuable insights into the biological functions of DEGs in glioma.