Researchers led by Chaemin Lim, Hyunsu An and Jihwan Park analyzed 2,596 long-read RNA-sequencing datasets to build a comprehensive, locus-specific map of transcribed transposable elements in human cells.

Long-read sequencing revealed transcript structures standard repeat annotations can miss, including transcripts combining multiple elements, alternative splice forms, revised gene boundaries and transposable-element–gene fusions.

The team identified 535,724 transposable-element-derived transcript models—roughly one in nine of the 4,891,349 elements cataloged by RepeatMasker—with about 94 percent represented by a single isoform.

In single-nucleus data covering 174,419 cells across eight GTEx tissues, transposable-element transcripts averaged 6.7 percent of the transcriptome and exceeded 10 percent in skeletal muscle and lung; LINE elements accounted for an average 58.5 percent of these transcripts.

The analysis also found 16,459 cell-type-differential transposable-element transcripts across 74 cell clusters, with especially high counts in skeletal muscle myocytes and alveolar cell types, suggesting links to cellular identity rather than sequencing depth alone.

The authors caution that their alignment approach may miss micro-exons shorter than 30 base pairs and recommend improved methods in future work; they intend the reference to support research on transposable elements in development, normal cell function and disease.