#37 · ML / AI methods

Databases annotate TE-initiated transcripts in cancer and enhancer perturbations

The dominant thread is curated, database-style resources that organize functional genomic data into queryable annotations. TEITbase reports transposable element-initiated transcripts in human cancers, cataloguing regulatory transcripts derived from TEs. ePerturbDB compiles experimental perturbation data for enhancers, which act as cis-regulatory elements controlling gene expression across developmental stages, external signalling, and cell states. A separate preprint characterizes highly conserved 3'UTR fragments across mammals using a transformer genomic foundation model and transfer learning. This work is methodologically distinct from the database entries and is computational rather than curated-resource oriented. What is new here is the extension of cancer-focused annotation to TE-initiated transcripts, alongside regulatory-element perturbation resources. These entries report no comparative benchmarks against one another, so relative performance or coverage cannot be assessed.