Prediction of Essential Genes in Comparison States Using Machine Learning

Area of Science:

• Computational Biology
• Genomics
• Machine Learning

Background:

• Identifying essential genes in comparison states (EGS) is crucial for understanding cellular processes, disease mechanisms, and drug development.
• Existing methods often struggle with unbalanced sample sizes and capturing dynamic network alterations.

Purpose of the Study:

• To introduce a novel machine learning method, Prediction of Essential Genes in Comparison States (PreEGS), for identifying EGS.
• To enhance prediction accuracy by incorporating topological and gene expression features and addressing sample imbalance.

Main Methods:

• PreEGS extracts a five-dimensional feature vector including topological and gene expression data for each gene.
• A positive sample expansion method is employed to handle unbalanced datasets.
• The random forests model (PreEGSRF) was selected for optimal performance after evaluating various classifiers.

Main Results:

• PreEGSRF demonstrated superior performance compared to six other methods in predicting EGS.
• On real datasets, PreEGSRF identified five essential genes and five KEGG pathways associated with leukemia.
• Predicted genes and pathways showed high consistency with previous studies and strong correlation with leukemia.

Conclusions:

• PreEGSRF offers high prediction accuracy and generalization ability for identifying essential genes.
• The method is broadly applicable for discovering disease-causing genes, cell fate drivers, and biological system biomarkers.