作者: Yasuyuki Ohkawa , Kazumitsu Maehara , Masatoshi Fujita , Takeru Fujii
DOI: 10.1101/2021.03.12.435089
关键词:
摘要: Statistical methods for detecting differences in individual gene expression are indispensable understanding cell types. However, conventional statistical have faced difficulties associated with the inflation of P-values because both large sample size and selection bias introduced by exploratory data analysis such as single-cell transcriptomics. Here, we propose concept discriminative feature cells (DFC), an alternative to using differentially expressed gene-based approaches. We implemented DFC logistic regression adaptive LASSO penalty perform binary classification discrimination a population interest variable obtain small subset defining genes. demonstrated that prioritized pairs non-independent artificial data, enabled characterize muscle satellite population. The results revealed well captured cell-type-specific markers, specific patterns, subcategories this may complement interpreting sets.