Automatic curation of LTR retrotransposon libraries from plant genomes through machine learning

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Transposable elements (TEs) are mobile DNA sequences vital for organismal adaptation and evolution.
• LTR retrotransposons are prevalent in plant genomes, comprising a significant portion of their DNA.
• Current methods for annotating TEs are manual, time-consuming, and require extensive bioinformatics expertise.

Purpose of the Study:

• To develop and validate a machine learning (ML)-based approach for the automated annotation of transposable elements in plant genomes.
• To significantly reduce the time and manual effort required for TE annotation compared to traditional methods.

Main Methods:

• A novel machine learning model was trained to identify and annotate transposable elements.
• The ML approach was applied to the genomes of four different plant species.
• Performance was evaluated using the F1-score and compared against conventional bioinformatics pipelines.

Main Results:

• The ML-based approach achieved a high F1-score of up to 91.18% across tested plant species.
• In one species (Oryza granulata), the ML method obtained a 93.6% F1-score in just 22.61 seconds, compared to approximately 6 hours for traditional methods.
• The results demonstrate a substantial acceleration and high accuracy of the proposed automated method.

Conclusions:

• Machine learning offers an efficient and accurate solution for automated transposable element annotation in plant genomes.
• This accelerated approach is highly suitable for large-scale sequencing projects and advancing plant genomics research.
• The developed ML method can aid in understanding phenotype variability, species evolution, and genome characteristics.