Using the nucleotide substitution rate matrix to detect horizontal gene transfer

Area of Science:

• Genomics
• Bioinformatics
• Evolutionary Biology

Background:

• Horizontal gene transfer (HGT) drives bacterial evolution and confers critical traits like antibiotic resistance.
• Existing HGT detection methods, relying on nucleotide composition or phylogenetics, suffer from high error rates.
• Improved HGT detection from sequence data is crucial for understanding bacterial adaptation and medical advancements.

Purpose of the Study:

• Introduce novel methods for detecting HGT based on nucleotide substitution rate changes.
• Enhance the accuracy and reduce the error rates of HGT detection compared to existing techniques.
• Investigate the utility of combining multiple HGT predictors for improved performance.

Main Methods:

• Developed new HGT detection methods analyzing changes in nucleotide substitution rates post-transfer.
• Utilized non-time-reversible models for enhanced HGT discrimination.
• Employed the random forests machine learning algorithm to combine multiple HGT predictors.

Main Results:

• New methods achieved up to 10 times lower error rates than GC content for simulated HGT events.
• Combining multiple predictors reduced the error rate from 38% to approximately 3%, an 18-fold improvement.
• Non-time-reversible models proved crucial for accurate HGT detection.

Conclusions:

• The novel HGT detection methods integrate phylogenetic and compositional approaches, offering significant improvements.
• These techniques provide order-of-magnitude gains in discriminating HGT events under diverse conditions.
• Combining multiple HGT indicators is essential for robust detection, aiding research into bacterial traits like antibiotic resistance and pathogenicity.