STEAP

Single cell Type Enrichment Analysis for Phenotypes (STEAP) uses scRNA-seq data and GWAS summary statistics to determine which cell-types are enriched in the GWAS phenotype. It is an extension to CELLECT and uses S-LDSC (Finucane et al., 2015), MAGMA (de Leeuw et al., 2015) and H-MAGMA (Sey et al., 2020) for enrichment analysis.

STEAP runs multiple post-processing steps on top of CELLECT output files:

Gene Set Enrichment Analysis (GSEA)
Cell-Type Correlation
Expression Specificity (ES) Gene Correlation

Installation

Step 1: Install (mini)conda

The pipeline works in conda enviroments, so you will have to install anaconda or miniconda first if not already on the system. You can do this by running:

wget "https://repo.anaconda.com/miniconda/Miniconda3-latest-Linux-x86.sh"
bash Miniconda3-latest-Linux-x86_64.sh
rm Miniconda3-latest-Linux-x86_64.sh

Step 2: Clone this repo and run the install script

First clone this repo, and then run the install.sh script by running:

git clone https://github.com/erwinerdem/STEAP.git
bash STEAP/install.sh
conda activate steap

This will create the necessary conda environments and clones the CELLECT repo and merges it with STEAP.

Getting Started

Any GWAS summary statistic of your interest can be used in this tutorial. Here, we will download and use the PGC depression GWAS as an example. Download can take a few minutes.

wget -O gwas/PGC_UKB_depression.txt https://datashare.is.ed.ac.uk/bitstream/handle/10283/3203/PGC_UKB_depression_genome-wide.txt --no-check-certificate

Munging GWAS

The summary statistic needs to be compatible with the pipeline. This is done by first munging the GWAS using the mtag_munge.py script from ldsc. This must be done in a special conda environment.

conda activate munge_ldsc

Then we can munge the PGC depression GWAS using:

python2 ldsc/mtag_munge.py \
--sumstats gwas/PGC_UKB_depression.txt \
--merge-alleles data/ldsc/w_hm3.snplist \
--a1 A1 \
--a2 A2 \
--snp MarkerName \
--p P \
--N-cas 170756 \
--N-con 329443 \
--signed-sumstats LogOR,0 \
--frq Freq \
--out gwas/PGC_UKB_depression

Each GWAS has different column names so change the parameters accordingly. More examples can be found in timshel-2020/src/prep-gwas_munge/README-cmds_munge_sumstats.txt. The mtag_munge.py script is also well documented. The output .sumstats.gz file will be used as input for the pipeline. Deactivating the munge_ldsc environment can be done simply using conda deactivate.

Converting scRNA Data to ES Matrix

Converting the scRNA to ES matrix requires CELLEX. This can be installed as mentioned in the CELLEX repository. Example notebooks converting the raw scRNA to ES matrices using CELLEX can be found in the cellex-notebooks.

Running the pipeline

To run the pipeline a config.yml must first be set up. For the PGC depression GWAS we will use the config.yml file. This file can be edited to instead include your own GWAS summary statistics or scRNA-seq data. Under SPECIFICITY_INPUT you can add the id and path of your ES matrix file. The id is simply the identifier for the dataset and the path the path to the file. Similarly, under GWAS_SUMSTATS you can add new munged GWAS summary statistics. If you want to include a different annotation file for H-MAGMA, you can edit this under HMAGMA_ANNOT. The results will be saved in the BASE_OUTPUT_DIR. For normal use it is recommended to only edit these parameters and leave the rest as is, unless you know what you are doing.

To run the enrichment analysis run:

snakemake --use-conda -j -s cellect-magma.snakefile --configfile config/config.yml
snakemake --use-conda -j -s cellect-h-magma.snakefile --configfile config/config.yml
snakemake --use-conda -j -s cellect-ldsc.snakefile --configfile config/config.yml

To instead run it on a Sun Grid Engine (SGE) cluster use:

qsub SGE_cluster_h-magma.sh
qsub SGE_cluster_ldsc.sh
qsub SGE_cluster_magma.sh

Running the post-processing notebook

To use the post-processing scripts just use the notebook after running the pipeline.

Another option is to use the STEAP post-processing appyter, where you can upload your prioritization.csv files which you get as output of the enrichment analysis.

Contact

Erwin Erdem (erwin.erdem@outlook.com)
Dr. Gennady Roshchupkin, PhD (g.roshchupkin@erasmusmc.nl)
Prof. dr. Steven Kushner, PhD (s.kushner@erasmusmc.nl)