Datasets used in the scECODA publication as Seurat objects in R .rds format

OverviewThis repository contains the processed single-cell RNA sequencing (scRNA-seq) datasets used in the study "Cell type composition drives patient stratification in single-cell RNA-seq cohorts" (Halter et al., 2026). The data was compiled to benchmark scECODA (single-cell Exploratory Compositional Data Analysis) alongside other state-of-the-art sample representation methods.Data ContentThe repository includes 11 patient cohorts comprising 697 individual samples across diverse biological conditions. The datasets are provided as R-format RDS files containing Seurat objects. Each object includes: Raw UMI Count MatricesCell-Type Annotations: High- and lo-granularity labels, including expert author-provided labels and automated annotations (HiTME/scATOMIC).Sample-Level Metadata: Cohort-specific clinical and biological variables such as disease status, treatment response (e.g., ICB response), age, and infection status (e.g., CMV, COVID-19), where applicable.Cohorts IncludedThe datasets span various tissues and pathologies, including:Cancer: Breast cancer (Bassez, Wu, Zhang), Lung adenocarcinoma (Kim), and Prostate cancer bone metastases (Kfoury).Infectious/Inflammatory: COVID-19 (Stephenson), Ulcerative Colitis (Smillie), and healthy donors stratified by CMV status (Gong & Sharma).Fibrosis: Pulmonary fibrosis (Adams).Batch effect: Sequencing technology (Joanito), healthy or COVID-19 samples from different datasets (Gong & Sharma, Stephenson, Zhu).Usage NotesThese objects are ready for immediate use in R for exploratory data analysis, patient stratification, and cell-type differential abundance testing. They serve as a standardized benchmark for testing new computational methods for cohort-level single-cell analysis.For the associated analysis code and the scECODA R package, please visit:https://github.com/carmonalab/ECODA_paper and https://github.com/carmonalab/scECODA, respectively.