ICIsAtlas 揭示泛癌免疫治疗图谱中的抑制性 NK 细胞生态位

Immune checkpoint inhibitors (ICIs) have reshaped the treatment in multiple tumors.

However, a substantial proportion of patients exhibit limited responses. The lack of harmonized, large-scale pan-cancer ICI datasets coupled with accessible analysis tools hinders the systematic discovery of response biomarkers and resistance mechanisms.

Therefore, we developed a comprehensive resource named ICIsAtlas, which encompassed curated transcriptomic and clinical data from 1,268 ICI-treated patients across eight tumor types, with an accompanying R package implementing a complete workflow for deconvolution and biomarker evaluation. Applying the ICIsAtlas framework, a systematic pan-cancer analysis was performed.

We identified the universal signatures that were specific for ICI response, including cooperative interactions among favorable immune cells.

In addition, we discovered a competitive cell community and SERPING1 + VEGFA+ Natural Killer (NK) cell-mediated immunosuppressive niche in non-responders, which were further validated with single-cell and multiplex immunohistochemistry data. The ICIsAtlas resource and R package represent a powerful, publicly available platform for hypothesis generation and biomarker discovery, which could be used to develop the next-generation biomarkers and therapeutic targets to improve tumor immunotherapy.