iPromoter-ET: Identifying promoters and their strength by extremely randomized trees-based feature selection

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Promoters are crucial DNA regions regulating gene transcription.
• Identifying promoters and their strength is vital in post-genomic research.
• Existing computational tools require enhancement to handle vast DNA sequence data.

Purpose of the Study:

• To develop an advanced computational model, iPromoter-ET, for identifying DNA promoter sequences.
• To enhance the classification of promoters into strong and weak categories.
• To provide an effective tool for promoter analysis in the era of big genomic data.

Main Methods:

• Feature extraction using k-mer nucleotide composition, binary encoding, and dinucleotide property matrix.
• Feature selection employing extremely randomized trees (extra trees).
• A two-layer model for promoter identification and strength classification, validated with Support Vector Machine and five-fold cross-validation.

Main Results:

• The iPromoter-ET model demonstrated superior performance in both promoter identification and strength classification.
• Achieved remarkable improvements in accuracy and stability compared to existing models.
• The model's effectiveness was validated through rigorous testing and comparison.

Conclusions:

• iPromoter-ET is a highly effective intelligent tool for identifying promoter regions and their strength.
• The model serves as a valuable complementary tool to existing methods in promoter analysis.
• Freely available datasets and codes facilitate broader adoption and further research.