LATTE for locus-specific quantification of transposable element expression across species

Transposable elements (TEs) are pivotal drivers of eukaryotic genome evolution and phenotypic diversity. However, their functional contributions to complex traits remain largely obscured by expression quantification challenges arising from high sequence homology and multi-mapping ambiguities. Here, we present LATTE, an efficient computational framework for defining and quantifying TE expression at locus-specific resolution by leveraging an innovative multi-indicator Expectation-Maximization (EM) algorithm. Extensive benchmarking against simulated datasets demonstrated that LATTE significantly outperformed existing state-of-the-art tools, achieving an accuracy of 0.998 at the subfamily level and 0.839 at the locus-specific level. Applying LATTE to 813 RNA-seq datasets across humans, cattle, and chickens, we quantified expression profiles of 2,703 TEs, followed by TE-expression quantitative trait loci (TE-eQTL) mapping. The colocalization rates between TE-eQTL and host gene-eQTL was low, revealing a distinct regulatory landscape of TE expression. This decoupled correlation between TEs and host genes are likely mediated by the differential expression of alternative transcripts. Through integrated TE-eQTL and genome-wide association studies on 3,746 complex traits across three species, we demonstrated that TEs constitute 204 (8.7%) additional associations with complex traits beyond gene-eQTL. More specifically, the Sjogrens syndrome-associated variant rs10954213 acts as a TE-eQTL that shifts the splicing landscape of IRF5, upregulating TE-containing transcripts while simultaneously suppressing canonical ones. Collectively, LATTE provides an efficient framework for studying TE expression across species, and our findings highlight the key role of TEs in understanding the genetic architecture of complex phenotypes.