ClassifyTE: a stacking-based prediction of hierarchical classification of transposable elements.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Transposable Elements (TEs), or jumping genes, are mobile DNA sequences influencing genome evolution.
• TEs can alter gene expression, increase mutation rates, and mediate large-scale genetic rearrangements.
• Accurate classification of TEs is crucial for understanding their impact on germline and somatic evolution.

Purpose of the Study:

• To develop an accurate and robust method for the hierarchical classification of Transposable Elements.
• To evaluate the performance of various machine learning techniques for TE classification.
• To provide an automated tool for classifying TEs up to the super-family level.

Main Methods:

• A stacking-based machine learning (ML) approach was employed for hierarchical TE classification.
• The proposed method, ClassifyTE, was trained and validated on multiple benchmark datasets.
• Performance was compared against existing state-of-the-art TE classification methods.

Main Results:

• The ClassifyTE system demonstrated superior performance, achieving significant average percentage improvements in hF measure across datasets.
• An automated, end-to-end hierarchical classification tool, ClassifyTE, was developed.
• High concordance was observed when evaluating ClassifyTE on a novel TE library, validating its accuracy at higher taxonomic levels.

Conclusions:

• ClassifyTE offers a highly accurate and automated solution for hierarchical TE classification.
• The tool facilitates a more precise analysis of the genetic and evolutionary roles of Transposable Elements.
• This advancement aids researchers in understanding genome dynamics and evolution driven by TEs.