Live-seq is a single-cell transcriptome profiling approach that use fluidic force microscopy to preserve cell viability during RNA extraction. Therefore, Live-seq can address a broad range of biological questions by transforming scRNA-seq from an end-point to a temporal analysis approach.
W. Chen, O. Guillaume-Gentil, et al., Genome-wide molecular recording using Live-seq, Nature, 2022
This repository contains the pipeline to be able to reproduce all figures of the paper.
00. Dependencies: Installs the dependencies for the project. You can find the package versions used in the paper in session_info.md
01. Preprocessing: Create a Seurat object from the count matrix generated after Live-seq and single-cell sequencing. Some basic single-cell QCs.
02. Live-seq: Subset the Seurat object to Live-seq cells only. Generate Live-seq-specific QCs. Downsampling of Live-seq.
03. scRNA-seq: Subset the Seurat object to single-cell RNA-seq (scRNA-seq) cells only. Generate scRNA-seq-specific QCs. Downsampling of scRNA-seq.
04. Live-seq scNRA-seq integration: Integration of both layers of data using Seurat's integration pipeline.
05. Live-seq with live cell imaging: Connecting the transcriptomic profile with a downstream phenotypic response, i.e. TNF upregulation upon LPS treatment
06. Clustering: Compute clustering accuracy (ARI, Barplot comparison and clustree)
07. Differential expression across cell types: Compute DE for each cluster (corresponding to a cell type/state) versus the rest for both Live-seq and scRNA-seq data. EnrichR on DE genes for BP GO terms and Mouse Cell Atlas, of each cluster versus Rest per sampling method
08. Analysis per cell type: Compute figures Live-seq manuscript per cell types: 1. tSNE per cell types colored by metadata and clustering, 2. tSNE colored per extracted volumes, 3. Avg expression per cell type - correlation between sc and live
09. Differential expression within cell type: Compare DE results obtained with scRNA and Live-seq. Identify GO Terms of genes detected only by live-seq or scRNA-seq to find any potential bias.
10. Downsampling scRNA-seq: Downsample scRNAseq data so that match complexity of Live-seq -> perform DE. Compare DE results obtained with scRNA and Live-seq. Downsampling scripts for scRNA-seq. (see 2. for downsampling of Live-seq data). Compare number of common genes between Live-seq vs scRNA-seq OR Live-seq vs Downsampled scRNA-seq.
./utils Some utility functions used across the scripts
This repository also contains the data used in the pipeline.
The main count matrix is downloadable on GEO: GSE141064. In the pipeline, it's automatically downloaded and parsed in the first script: 01. Preprocessing
- mouseGeneTable87_mCherry_EGFP.txt and Mus_musculus.GRCm38.100_data.annot.txt: The gene annotation information (gene name, biotype, exon length, ...)
- gene.blacklist.csv: List of 20 genes that are blacklisted and removed from many analyses, which are derived from the 0 pg input RNA negative control.
- InfoContent_Updated_VX-ASPC-9_4.csv and meta.final.csv: Sample (cell) metadata
- log.foreach.txt: Log of clustering script
- s.genes.mouse.rds and g2m.genes.mouse.rds: Cell cycle genes used to assign cell cycle scores