Statescope: a comprehensive framework designed to capture cell states

Malignant and non-malignant cells in the tumor microenvironment adopt diverse functional states that collectively shape cancer progression and treatment response. Deconvolution of bulk RNA-seq profiles have offered a powerful means to determine cell type fractions at scale. To enable an accurate cell state analysis, we established a Bayesian integrative deconvolution framework tailored for tumor bespoke analysis by integrating malignant cell purity with the RNA-seq data.
We present Statescope, a comprehensive framework designed to capture cell states. Statescope enables multi-omics deconvolution by integrating cell fractions derived from DNA data, followed by capturing inter-sample variations in cell type-specific gene expression profiles and cell states. Using bulk RNA-seq datasets with known gene expression profiles and cell fractions, as well as simulated bulk RNA-seq data from multiple scRNA-seq datasets, we systematically evaluated the Statescope framework, demonstrating its ability to correctly capture cell states for both malignant and non-malignant cells in tumor samples. Applying Statescope to The Cancer Genome Atlas (TCGA) cohorts, we identified previously recognized cell states including neutrophil subpopulations in non-small cell lung cancer (NSCLC) and mixed tumor cell states in pancreatic ductal adenocarcinoma (PDAC). Statescope is able to transfer discovered cell states to new single cell or bulk RNA-seq data. We transferred cell states discovered in the TCGA-NSCLC cohort to the randomized phase 2 POPLAR and phase 3 OAK immunotherapy trial cohorts in which we identified individual and combined cell states that can serve as a predictive biomarker for immunotherapy response