Using weighted features to predict recombination hotspots in Saccharomyces cerevisiae

Area of Science:

• Genetics
• Bioinformatics
• Computational Biology

Background:

• Accurate prediction of recombination hotspots and coldspots is vital for understanding recombination mechanisms.
• Existing models show promise but require further improvement for enhanced predictive power.

Purpose of the Study:

• To develop a novel DNA sequence classifier for predicting recombination hotspots and coldspots.
• To incorporate k-mer frequency and DNA physical-thermodynamic properties as weighted features.

Main Methods:

• A novel classifier was developed using k-mer frequency, physical, and thermodynamic properties of DNA sequences.
• Weighted features were incorporated into the classifier model.
• The classifier was applied to both ORF and non-ORF sequences in Saccharomyces cerevisiae.

Main Results:

• The classifier achieved 90% accuracy in predicting recombination hot/cold spots in Saccharomyces cerevisiae ORFs.
• This represents a ~5% improvement over existing methods like iRSpot-PseDNC, IDQD, and Random Forest.
• High accuracy was also observed in predicting non-ORF recombination hot/cold spots.

Conclusions:

• The novel classifier demonstrates superior performance in identifying recombination hotspots and coldspots.
• The model's feature incorporation strategy offers a significant advancement in predictive accuracy.
• The approach shows broad applicability for sequence classification tasks in bioinformatics and genetics.