Abstract Despite improved survival for many cancer patients treated with immune checkpoint inhibitors (ICIs), interindividual responses vary greatly. Identifying patients likely to benefit from ICI is therefore crucial. We investigated the tumor microenvironment (TME) of advanced clear cell renal cell carcinoma (ccRCC) by performing an extensive relabeling of cell types from single-cell RNA-seq data validated for use in cell deconvolution of bulk RNA-seq data. We leveraged this approach to conduct a comprehensive analysis of predicted cell type fractions within the TME of ccRCC patient cohorts enrolled in ICI clinical trials. We found distinct tumor and immune cell composition across metastatic sites associated with ICI treatment response. Stratification of patients based on TME revealed three distinct subtypes characterized by varying compositions of CD8+ T cells and plasma B cells. These subtypes were associated with differential responses to PD1 blockade. Based on these findings, we further developed a Tumor-Immune Differential (TID) score that accurately predicts response to ICI (AUC-ROC=0.88, log-rank tests for PFS P 10Gb). Only the processed data to reproduce the downstream analysis are in the data/ folder. In order to reproduce any result and figure, you should start with the Jupyter notebook "5.a.*". However, if you want to reproduce the raw data processing, the following Jupyter notebooks with file names starting with the numbers "1, 2, 3, 4" and "7, 8" were used to download, filter, harmonize and merge the methyl-array and gene expression data, respectively. Other notebooks will be able to run without this process. SingularityImage.sif To seamlessly reproduce the entire workflow of Jupyter notebooks without dependency issues, you can use the Singularity image. To open a Jupyter Lab session and execute the notebooks from the CodeNotebook directory (uncompressed) mounted inside the Singularity image, please run this command: singularity exec --bind /pathTo/CodeNotebook/:/home/pathInside/mnt SingularityImage.sif jupyter lab --no-browser--NotebookApp.iopub_data_rate_limit=1.0e10 Please, note that due to dependencies between the notebook results, you should follow the numbering to run each notebook. So, run the "5.a.*", then the "5.b.*", and so on. For any issues and questions, please contact Florian Jeanneret and Christophe Battail.