Background: Human cancer cell lines are fundamental models for cancer research and therapeutic strategy development. However, there is no characterization of circular RNAs (circRNAs) in a large number of cancer cell lines. Methods: Here, we apply four circRNA identification algorithms to heuristically characterize the expression landscape of circRNAs across ~ 1000 human cancer cell lines from CCLE polyA-enriched RNA-seq data. By using integrative analysis and experimental approaches, we explore the expression landscape, biogenesis, functional consequences, and drug response of circRNAs across different cancer lineages. Results: We revealed highly lineage-specific expression patterns of circRNAs, suggesting that circRNAs may be powerful diagnostic and/or prognostic markers in cancer treatment. We also identified key genes involved in circRNA biogenesis and confirmed that TGF-