Identify essential genes based on clustering based synthetic minority oversampling technique

Area of Science:

• Synthetic biology
• Genomics
• Bioinformatics

Background:

• Essential gene prediction is vital for synthetic biology.
• Current sequence-based predictors need performance improvement.
• Imbalanced datasets (fewer essential than non-essential genes) hinder predictor accuracy.

Purpose of the Study:

• To develop a novel computational method for accurate essential gene prediction.
• To address the challenge of imbalanced training data in essential gene identification.
• To enhance the extraction of sequence-based features for discriminating essential genes.

Main Methods:

• Proposed a new over-sampling strategy: Clustering based Synthetic Minority Oversampling Technique (CSMOTE).
• Developed a protocol named iEsGene-CSMOTE, integrating CSMOTE with Z curve and global sequence features.
• Utilized Support Vector Machines (SVM) for classification.

Main Results:

• iEsGene-CSMOTE demonstrated superior performance compared to existing methods via rigorous jackknife cross-validation.
• The proposed method significantly improved sensitivity (Sn) by 35.48% and balanced accuracy (BACC) by 11.25% over the λ-interval Z curve method.

Conclusions:

• The iEsGene-CSMOTE protocol effectively overcomes the imbalanced data issue in essential gene prediction.
• This novel approach offers a significant advancement in computational identification of essential genes.
• The findings have implications for improving gene function annotation and synthetic biology applications.