scRNA-seq expression matrices for benchmarking single-cell genome-scale metabolic model construction

These data support the study “A benchmarking framework for single-cell genome-scale metabolic model construction.” This record contains the quality-controlled scRNA-seq expression matrices used to benchmark 26 strategies for single-cell genome-scale metabolic model construction across nine datasets. It includes complete datasets and randomly sampled subsets, each provided as quality-controlled raw expression matrices (R) and matrices normalized using Linnorm and imputed using SAVER (LS). Here, R denotes expression data after quality control but before normalization and imputation, rather than raw sequencing reads. The LUAD archive additionally contains expression matrices generated by perturbing the randomly sampled LS dataset to achieve target Spearman rank correlations with the unperturbed matrix. These matrices were used for sensitivity analysis. Code repository: https://github.com/ChenYuGroup/scGEM_benchmark