Cell deconvolution predicts the cellular heterogeneity in the ccRCC tumor microenvironment

Cell deconvolution methods have emerged in recent years as relevant bioinformatics approaches for predicting the proportions of cell types present in biological samples profiled by bulk RNA-seq. Within the framework of the European KATY project (https://katy-project.eu/), we are interested in the heterogeneity of the tumor microenvironment of the clear cell renal cell carcinoma (ccRCC) and its influence on the ability to predict a patient's response to a treatment.
To meet this objective, we have optimized a bioinformatics protocol for the analysis of single-cell RNA-seq data and the prediction of cell fractions by deconvolution methods. We evaluated the cell deconvolution methods
CIBERSORTx and MuSiC. The single cell RNA-seq matrix used in our work to perform cell deconvolution was optimized from the one obtained from 11 adult patients with ccRCC.
We generated pseudo-Bulk RNA-seq matrices from the single cell RNA- seq matrix to assess the performance of cell deconvolution methods. Then, we performed deconvolution on a cohort of bulk RNA-seq data consisting of 311 ccRCC tumor samples and assessed the quality of our predictions by comparison with tumor purity scores