sc-TWAS

References

Mapping the landscape of lineage-specific dynamic regulation of gene expression using single-cell transcriptomics and application to genetics of complex disease

Hanna Abe¹,Phillip Lin², Dan Zhou², Douglas Ruderfer^2,3, Eric R. Gamazon^2,3

¹Vanderbilt University, Nashville, TN, USA
²Division of Genetic Medicine, Department of Medicine, Vanderbilt University Medical Center, Nashville, TN, USA
³Vanderbit Genetics Institute, Vanderbilt University Medical Center, Nashville, TN, USA
⁴Clare Hall, University of Cambridge, Cambridge, United Kingdom

Send correspondence to:
Eric R. Gamazon ericgamazon@gmail.com

Code being maintained by:
Hanna Abe abehanna1@gmail.com

Using this resource

Download prediction models

The cell type and cell state adjusted prediction models are available on Zenodo at (to be provided up on publication). Pre-trained tissue-specific PrediXcan gene expression models leveraged here are available for download from the JTI repository (https://doi.org/10.5281/zenodo.3842289).

Application of models to GWAS summary data

To apply the cell type prediction models to a GWAS summary statistics and get TWAS results, we use a method that has been developed by Barbeira et al, 2018, S-PrediXcan (https://github.com/hakyimlab/MetaXcan/blob/master/software/SPrediXcan.py).
Reference: Barbeira, Alvaro N., et al. "Exploring the phenotypic consequences of tissue specific gene expression variation inferred from GWAS summary statistics." Nature communications 9.1 (2018): 1-20. (https://www.nature.com/articles/s41467-018-03621-1)

PheWAS using cell type prediction model application to UKBB

We have applied our cell type models to > 1000 GWAS summary statistics in the UKBB. The R package below performs PheWAS using the UKBB associaitions. Given a gene or multiple genes, the package will return signficant associations (those that pass the multiple testing correction) and returns a PheWAS plot. If the gene is not imputed in any cell type model, it will not return any output. Please refer to the src folder above to see details about package and list of available iGenes and cell type models.

TUTORIAL

Install the R package.

R CMD INSTALL sctwas_0.3.0_R_x86_64-pc-linux-gnu.tar.gz

library(sctwas)

# First get the table of  the association
data <- getGraphDatav2(geneNames=c("CHRNA3", "ARL17B", "HYI"), cellTypeId= "FPP", timePoint="D11")

# We then graph the PheWAS manhattan plot using the data table above
phewas_manhattan(data, cellTypeId="FPP", timePoint="D11")

The function above returned the following figure. The x-axis shows the class of phenotypes in the UKBB, y-axis is -log10(TWAS pvalue). The function labels only those associations that pass the multiple testing treshold. This figure shows the results from the application of the FPP cell type model to the GWAS summary statistics.