/li_et_al_cell_2018_melanoma_scrna

Primary LanguageR

Single-cell RNA analysis of immune cells from melanoma tumors

This repository holds the code that reproduce the analysis done in the Li et al, 2018 paper. It also downloads the required processed and auxilary data.

The core analysis is done with the metacell R package.

Quick Links

  • Metacell paper: Baran et al. 2018 (bioarxiv).
  • Metacell R package homepage (with functions reference and usage vignettes).
  • Raw data is available under EGA accession EGAS00001003363 (access is restricted, follow guidelines at the EGA site)
  • Processed data is avaliable under GEO accession GSE123139
  • Raw UMI counts for all cells with metacell membership information: Li2018_umi_counts.tar.gz. This gzipped archive contains a gene-cells UMI table in matrix-market format and a metadata table per cell. The metadata table also contains the metacell membership and the annotated group per metacell model - all cells, T/NK cells and non-T/NK cells.
  • Code repository for the TCR sequences analysis is available at TCRseq github page.

Requirements

R with these packages:

  • Matrix
  • pheatmap
  • flowCore
  • dplyr
  • glmnet
  • plotrix
  • MASS
  • data.table
  • metacell

Note: Metacell is implemented in R and C++. In particular it uses the Tanay group tgstat library that utilizes shared memory and distributed computing (as well as some specific optional CPU features). The package is tested on linux and macbooks, and is currently not compatible on Windows. A typical application of metacell requires at least 16G RAM. For the current dataset we recommend a dual CPU multi-core workstation with 128GM RAM or more.

Usage

In an R session opened on the repository root directory

# Loading code and downloading required data files
source("pipe.r")

# Building the metacells
build_metacells()

# Generate figures
generate_figs()

# Reproduce the Guo et al 2018 (lung cancer scRNA data) analysis
source("guo2018.r")

Note: Metacell generation was sensitive to the initial random seed, so the exact metacell solution cannot be exactly reproduced. build_metacells function is included for reference, to see how the metacells were produced. In order to reproduce the exact figures as in the paper (e.g. supporting the manually selected metacell IDs for annotation), the metacell object used in the paper is supplied and used by the generate_figs function. This function also does a cross-metacell comparison to compare the cell membership between the 2 metacell solutions, which demonstrate that the differences are not substantial.

Contact

For help, please contact yaniv.lubling@weizmann.ac.il