panHiTE: a comprehensive and accurate pipeline for TE detection in large-scale population genomes

Transposable elements (TEs) are key drivers of genomic variation and species evolution. Advances in high-throughput sequencing have enabled whole-genome sequencing of individuals or subspecies, facilitating the identification of population-specific variations. Detecting population-specific TE insertions at scale is crucial for understanding species-specific phenotypic traits. However, tools for constructing comprehensive pan-TE databases remain limited. To address this gap, we develop panHiTE, a population-scale TE detection and annotation tool with several core innovations. panHiTE features a deep learning-based long terminal repeat retrotransposon (LTR-RT) detection algorithm, outperforming existing tools in both sensitivity and precision. It also introduces a novel de-redundancy algorithm, which eliminates highly divergent redundant TE instances, significantly reducing the size of the TE library. Additionally, panHiTE can detect low-copy TEs, which are overlooked in individual genome analyses and absent from existing databases due to their rarity. Furthermore, panHiTE allows for TE-gene association analysis, enabling comprehensive insights into TE-driven phenotypic variation. panHiTE, powered by a Nextflow pipeline, enables efficient and scalable TE detection in large plant genomes and has successfully been applied to hundreds of plant population genomes, demonstrating its effectiveness and scalability.