Pan-cancer single-cell RNA-seq identifies recurring programs of cellular heterogeneity

This resource provides the R code to reproduce key results described in Kinker, G.S., et al. Pan-cancer single-cell RNA-seq identifies recurring programs of cellular heterogeneity. Nat Genet 52, 1208–1218 (2020).

The analyses are divided into 6 main modules:
1. Identifying discrete and continuous patterns of expression heterogeneity within cell lines and human tumors.
2. Definying heterogeneity patterns that are shared between multiple cell lines and between multiple human tumors (i.e. recurrent heterogeneous programs, RHPs).
3-4. Comparing RHPs found in cell lines to RHPs found in human tumor samples.
5-6. Evaluating the association between expression and genetic heterogeneity in cell lines.

For questions, please contact gabriela.kinker@gmail.com

Getting started

1. Clone Github repository.

git clone https://github.com/gabrielakinker/CCLE_heterogeneity.git

2. Set the working directory to CCLE_heterogeneity.

3. Download the data provided (CCLE_scRNAseq_github.zip), which includes processed, post-QC scRNA-seq data, into the CCLE_heterogeneity GitHub directory.

4. Unzip files.

unzip CCLE_scRNAseq_github.zip && mv CCLE_scRNAseq_github/* . && rm -d CCLE_scRNAseq_github

5. Install required R packages.

Rscript packages.R

5. Run one of the 6 code modules in R.

• module1_expr_heteroge.R
• module2_rhp.R
• module3_vitro_vs_vivoI.R
• module4_vitro_vs_vivoII.R
• module5_cna_subclones.R
• module6_subclones_vs_expr_heteroge.R

6. Output files will be saved in the Output directory.

General notes

• Please, note that due to the stochastic nature of methods such as T-distributed Stochastic Neighbor Embedding (t-SNE) and nonnegative matrix factorization (NMF) outputs from module1_expr_heteroge.r might slightly differ dependending on the version of R/R packages used.

• As all outputs are already provided in the Expected_results directory (CCLE_scRNAseq_github.zip), it is possible to run each code module independently.

• Running module1 takes several hours.

• Code used for quality control and processing of UMI counts (UMIcount_data.txt) can be found in the additional_code directory.

Requirements

• R (tested in version 3.6.3).