Downstreamer for single‐cell gene prioritization

The following steps outline the process of prioritizing genes using Downstreamer to create single-cell gene regulatory networks. Downstreamer integrates GWAS summary statistics with a correlation matrix containing the co-regulation patterns within specific cells to identify key genes potentially contributing to disease development.

Preceding the steps outlined here, a gene by gene correlation matrix was generated by calculating the weighted Pearson correlation coefficients of the gene-expression for each gene-pair combination. This matrix was then decomposed into its eigenvalues and eigenvectors, and the eigenvectors that explained 80% were selected. The selected eigenvectors were then reformatted into a format that can be used as input for Downstreamer.

1. GWAS gene aggregation

In this step the GWAS summary statistics are converted from p-values per variant to an aggregate p-value per gene, accounting for LD structures (1000 Genomes phase 3) using PascalX (version 0.0.3).

Arguments used:

Argument	Description
--refpanel	Reference panel path	1kGP_high_coverage_Illumina
--gwas	GWAS summary statistics path	{gwas}.snppval.txt.gz
--annotation	Genome annotation file	genes_Ensembl94.txt
--threads	Number of threads for scoring	20
--outfile	Output file	{gwas}.txt
--rscol	Rs ids column index	0
--pscol	P-value column index	1
--window	Gene window	25000

2. Downstreamer

Next, the aggregated p-values and reformated eigenvectors are input into Downstreamer to prioritize key genes.

Arguments used:

Argument	Description
--mode	Downstreamer mode	ENRICH
--gwas	GWAS gene p-values	{gwas}.txt
--geneCorrelations	Gene expression matrix force normalized and split per chromosome	permutationGeneCor/geneCorForceNormalchr_
--output	Output file path	{gwas}_keygenes_covCor
--genes	File with gene information col1: geneName (ensg) col2: chr col3: startPos col4: stopPos col5: geneType col6: chrArm	genes_Ensembl94_protein_coding.txt
--expressionEigenVectors	Selected eigenvectors of gene-genecorrelation matrix	{eigenvectors}
--covariates	Corrects for potential inflation in correlation signal using median signal	medianGwasSignal.txt
--eh	Exclude HLA locus during pathway enrichment (chr6 20mb - 40mb)
-t	Maximum number of calculation threads	8
--forceNormalGenePvalues	Force normal gene p-values before pathway enrichment