Critical evaluation of damaged cell quality control in single-cell RNA sequencing [Liver Case Study: Integrated filtered output]

Mouse liver dataset (GSE221481) containing four samples, underwent filtering by a damaged cell detection strategy before integration and cell type annotation. Sixteen filtering strategies were implemented, with the integrated output available in the following .rds files. These include eight tool-based strategies: DamageDetective, ddqc, DropletQC, miQC, SampleQC, scater (filter and PCA), and valiDrops. As well as eight manual strategies, including fixed thresholds (fixed_mito_5, fixed_mito_10, fixed_mito_25) and adaptive thresholds set using MAD (3+) outlier detection (mito, mito_ribo, mito_features, malat1, malat1_features).小鼠肝脏数据集（GSE221481）包含4个样本，在开展整合分析与细胞类型注释前，已通过受损细胞检测策略完成过滤流程。本研究共实施16种过滤策略，整合后的输出结果可通过以下.rds文件获取：其中8种为基于工具的过滤策略，分别为DamageDetective、ddqc、DropletQC、miQC、SampleQC、scater（过滤与主成分分析）及valiDrops；剩余8种为手动配置的过滤策略，涵盖固定阈值策略（"fixed_mito_5"、"fixed_mito_10"、"fixed_mito_25"），以及采用MAD（中位数绝对偏差，Median Absolute Deviation，3倍离群值检测阈值）开展异常值检测的自适应阈值策略：mito、mito_ribo、mito_features、malat1、malat1_features。