Genomic object detection: An improved approach for transposable elements detection and classification using...

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Transposable elements (TEs) analysis is vital for understanding eukaryotic genomes.
• Current TE detection and classification methods are complex and time-consuming.
• Machine learning (ML) has been applied to TE classification, but not extensively for detection.

Purpose of the Study:

• To introduce YORO, a novel detection-classification strategy for transposable elements in genomic sequences.
• To adapt a computer vision convolutional neural network (YOLO) for genomic object detection.
• To enable prediction of genomic object position, length, and classification in large DNA sequences.

Main Methods:

• Developed YORO, a strategy based on convolutional neural networks adapted from computer vision.
• Trained the neural network using internal protein-coding domains of LTR-retrotransposons.
• Evaluated performance using metrics such as precision, recall, accuracy, F1-score, and execution times.

Main Results:

• YORO successfully detects and classifies genomic objects, specifically LTR-retrotransposons.
• The method demonstrates high performance in terms of accuracy and efficiency.
• Execution times and time ratios indicate significant improvements over existing methods.

Conclusions:

• YORO represents a promising deep learning approach for the next generation of genomic analysis tools.
• The strategy enables efficient detection and classification of transposable elements in large DNA sequences.
• This work facilitates more comprehensive and rapid analysis of eukaryotic genomes.