scTHREAD: a cell-resolved long-read transcriptome database with linked isoform, poly(A), junction, and allelic evidence

Cell-resolved transcript structures are essential for understanding how isoform usage, alternative polyadenylation, and allelic expression vary across cell types. However, existing single-cell long-read datasets remain fragmented across studies, platforms, species, and workflows, hindering comparison and reuse. Here, we present scTHREAD (single-cell Transcriptome Hub for Long-Read Evidence, Annotation, and Discovery; https://scthread.ai4sc.ac.cn/), an open-access database for integrating, exploring, and retrieving single-cell long-read transcriptomic evidence. It comprises 453 runs from 34 human and mouse datasets, covering 923 389 cells profiled by Oxford Nanopore and PacBio. A uniform, platform-aware workflow catalogues 12 471 531 run–transcript observations, of which 30% carry one of 210 341 distinct Ensembl transcript identifiers. Isoform and poly(A)-site usage, splice-junction support, and allele-aware counts remain linked to their source study, run, cell type, and single cell, ensuring full provenance, and differential usage is precomputed where coverage permits. Records are explored through a web interface and retrieved through bulk downloads or a documented API. scTHREAD recapitulates known cell-type-specific isoform patterns and identifies reproducible monocyte–T-cell differences in CCND3, CUTA, and PSMB8 junction usage across three cohorts and two sequencing technologies, establishing it as a traceable resource for investigating RNA processing and transcript regulation at isoform resolution.