Code script of our experiments.

Single-cell RNA sequencing (scRNA-seq) is a high-throughput experimental technique for studying gene expression at the single-cell level. As a key component of single-cell data analysis, differential expression analysis (DEA) serves as the foundation for all subsequent secondary studies. Despite the fact that biological replicates are of vital importance in DEA process, small biological replication is still common in sequencing experiment now, which may impose problems to current DEA methods. Therefore, it is necessary to conduct a thorough comparison of various DEA approaches under small biological replications. Here, we compare 6 performance metrics on both simulated and real scRNA-seq datasets to assess the adaptability of 8 DEA approaches, with a particular emphasis on how well they function under small biological replications. Our findings suggest that DEA algorithms extended from bulk RNA-seq are still competitive under small biological replicate conditions, whereas the newly developed method DEF-scRNA-seq which is based on information entropy offers significant advantages. Our research not only provides appropriate suggestions for selecting DEA methods under different conditions, but also emphasizes the application value of machine learning algorithms in this field.单细胞RNA测序（scRNA-seq）是一种用于在单细胞水平开展基因表达研究的高通量实验技术。作为单细胞数据分析的核心组成部分，差异表达分析（DEA）是所有后续衍生研究的基础。尽管生物学重复在DEA流程中至关重要，但当前测序实验中生物学重复样本量偏小的情况仍较为普遍，这会给现有DEA方法带来诸多挑战。因此，在低生物学重复条件下对各类DEA方法进行全面比较具有重要意义。本研究基于模拟数据集与真实scRNA-seq数据集，对比了6项性能指标，以评估8种DEA方法的适配性，重点探究其在低生物学重复场景下的表现能力。研究结果表明，源自批量RNA测序（bulk RNA-seq）的DEA算法在低生物学重复条件下仍具备竞争力，而基于信息熵开发的新型方法DEF-scRNA-seq则展现出显著优势。本研究不仅为不同应用场景下DEA方法的选择提供了合理建议，同时也凸显了机器学习算法在该领域的应用价值。