Leveraging long-read assemblies and machine learning to enhance short-read transposable element detection and...

Area of Science:

• Genomics
• Bioinformatics
• Molecular Evolution

Background:

• Transposable elements (TEs) are mobile genetic sequences crucial for genome evolution.
• Long-read sequencing enhances TE detection accuracy but is costly.
• Short-read sequencing methods show limitations in accurately detecting TEs from real data.

Purpose of the Study:

• To develop a machine learning approach (TEforest) for accurate TE insertion and deletion discovery and genotyping using short-read sequencing data.
• To leverage TEs identified via long-read sequencing as training data for the machine learning model.

Main Methods:

• Utilized a sensitive algorithm to identify potential TE insertion/deletion sites from short-read alignments.
• Extracted relevant features from short-read alignments for TE detection.
• Trained a random forest model using a ground-truth dataset of TEs identified by long-read sequencing.

Main Results:

• TEforest demonstrated superior performance compared to traditional methods, identifying more true positives and fewer false positives.
• The method accurately infers genotypes and precise insertion breakpoints across various read lengths and coverages.
• TEforest effectively learns short-read signatures of TEs previously only detectable with long reads.

Conclusions:

• TEforest bridges the gap between large-scale population studies and high-accuracy long-read assemblies for TE analysis.
• This user-friendly tool facilitates the study of TE prevalence and phenotypic effects across genomes.