Abstract Large single-cell atlases can serve as references for the analysis of smaller-scale studies. However, there are strong biases among atlases due to their varying sample sources, conditions, cell type annotation strategies and resolutions. Selecting appropriate atlas for future or smaller-scale studies is complicated due to unclear understanding of atlases, that limits the future or smaller-scale studies take advantage of the wealth of data that such atlases provide. For example, the tumor-infiltrating immune cell atlases including pan-cancer and cancer type-specific atlases provide comprehensive profiling of immune cells, which desperately facilitates future single-cell studies. However, lack of a comprehensive assessment and comparison of tumor immune cell atlases with standardized benchmarks blocked the application of the most appropriate atlas for smaller-scale study. In this study, we compared pan-cancer immune cell atlases including TICA, pan-cancer T cells and pan-cancer blueprint atlases, and cancer-type specific immune cell atlases containing ABTC, HCC and STAD atlases. We quantified the correlations of cell types and observed similar heterogeneity among atlases and unique immune cell populations in TICA as well. Then, we assessed the performance of atlases on accurately annotating immune cell types in the widely studied PBMC datasets—in which cell types were well annotated. Comparative analysis revealed the complementarity within atlases on mapping success, accuracy, clustering, annotatability and stability by benchmarking reference-based populations to reference-free de novo clustering on real immunotherapy datasets. We identified potential indicators of immune response and found that reference-based populations outperformed reference-free clusters in recognizing immunotherapy-related cell states. Our study can be easily extended to future atlases. Citation Format: Jing Yang, Qi Liu, Yu Shyr. A scalable comparison of tumor immune atlases [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2024; Part 1 (Regular Abstracts); 2024 Apr 5-10; San Diego, CA. Philadelphia (PA): AACR; Cancer Res 2024;84(6_Suppl):Abstract nr 2261.
Read full abstract