MLDEG: A Machine Learning Approach to Identify Differentially Expressed Genes Using Network Property and Network...

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Identifying differentially expressed genes (DEGs) is crucial for transcriptome data analysis.
• Existing DEG detection methods show inconsistent performance across different datasets and traits.
• Determining appropriate significance cutoffs remains a challenge in DEG analysis.

Purpose of the Study:

• To develop a robust ensemble model for accurate DEG identification.
• To refine DEG candidates by incorporating network information.
• To overcome the limitations of existing DEG detection methods.

Main Methods:

• Developed an ensemble model integrating network propagation and network properties.
• Re-classified DEG candidates with weak evidence from existing tools.
• Utilized 10 RNA-sequencing datasets from the Gene Expression Omnibus (GEO).

Main Results:

• The ensemble model achieved superior performance in identifying ground truth genes across 10 RNA-seq datasets.
• The method ranked first in detecting ground truth genes in eight out of ten datasets.
• Existing DEG methods exhibited significant performance variations across the tested datasets.

Conclusions:

• The proposed ensemble model significantly improves DEG identification accuracy in transcriptome data.
• The method's ability to integrate network information offers a more robust approach to DEG analysis.
• The model's design allows for the natural incorporation of new DEG detection methods.