Single Cell/Nuclei RNA-Seq
UM Bioinformatics Core
2023-08-29
Methods
Raw sequence data is transformed to gene-barcode count matrices using CellRanger (Zheng et al., 2017). Further data analysis is primarily performed using the Seurat package (Stuart et al., 2019). Gene-barcode matrices and metadata for each sample are loaded and further filtering and clustering analyses were performed as described in the Seurat tutorials (Seurat tutorials). Aberrant cells are filtered (low complexity, duplets, or apoptotic cells) and based on detected debris/contamination DecontX (Yang et al., 2020) may be run. Counts are normalized using the default normalization approach and variable features were identified. Where appropriate, anchor points were then generated across related datasets and used for SCTransform data integration. Principal component analysis (PCA) is then performed to identify significant PCA components used to find nearest neighbors followed by graph-based, semi- unsupervised clustering into distinct populations. Projections are generated using uniform manifold approximation (Becht et al., 2018) and marker genes are identified through differential gene expression pairwise comparisons (Wilcoxon rank-sum test for single-cell gene expression; FindAllMarkers function). Cell-type predictions were also generated with scCATCH (Shao et al., 2020). Suitability and approach for trajectory analysis is determined based on experimental design and disease model (Saelens et al., 2019; Lange et al., 2022; Cao et al., 2019; Wolf et al., 2019; La Manno et al., 2018; Bergen et al., 2020).