CONTENT

A method for multi-context genetic modeling of transcription regulation

Currently included in the repository is the software for running CONTENT as well as the context-by-context (classic TWAS/predXcan) approach. The implementation of both methods takes advantage of fast implementations of penalized linear models through package bigstatsr[1]. Future plans for development include a pipeline and instructions for conducting your own TWAS with hierarchical FDR[2,3]. In the meantime, feel free to reach out to us at mjthompson (at) ucla (dot) edu if you would like us to help you with your own TWAS using CONTENT and/or the context-by-context approach as well as on GTEx or our single-cell PBMC RNAseq dataset.

Running CONTENT and the context-by-context approach

CONTENT is run as an Rscript via command line, one run per gene (and all its measured contexts). This script will generate weights and cross-validated predictors for both the context-by-context approach and CONTENT, assuming there are at least 3 contexts for the CONTENT approach to be identifiable. The required arguments are as follows:

• --X_file A tab-delimited genotype matrix with the set of individuals in all Y_files and their corresponding cis-SNPs. Row names are individuals, no column names.
• --Y_file_dir Directory containing only the tab-delimited expression files for each context. The individuals (row names) must be a subset of those in the genotype (X_file). Contains an unnamed column with the expression---or if cov_file_dir is NULL, the residual expression for a given context.
• --cov_file_dir Specify this directory if you want to residualize the expression. The covariates are context-specific and should share the same name as the expression files such that file $i$ of the expression corresponds to file $i$ of the covariates. E.g. "$Y_file_dir/Liver" and "$cov_file_dir/Liver". If NULL, CONTENT will run assuming the expression has already been residualized. Needs to (1) be the same row-length as the corresponding expression file (2) contain the same rownames (3) contain no column names (4) use a tab delimiter.
• --out_dir Where do save the cross-validated predictors and their corresponding performance statistics. Use the same directory for all genes, and it will contain the predictors as well as their pvalues and $R^2$ values for each gene for convenient reporting.
• --snps A tab-delimited txt file containing information from your .bed or other genotype file. Contains 6 columns and the number of rows corresponds to the number of snps. No column names or row names. Col2 must be rsIDs (or matching your GWAS summary statistic SNP names) in order for FUSION/TWAS to work
  1. chromosome
  2. rsID
  3. location CM (not exactly necessary for TWAS)
  4. location/locus on chromosome
  5. allele1
  6. allele2
• --twas_dir Where to save the TWAS weights. We recommend using the same directory for all genes and it will conveniently store the results (such that all significantly cross-validated predicted genes will have their weights in the same directory and we can point FUSION/TWAS to this single directory).
• --gene_name Add this prefix to all the saved results files, this is necessary to distinguish your results.

There are also additional arguments that one can change:

• --verboseShow progress and output as the script runs.
• --signif Manually override the default (0.1) nominal prediction pvalue of when a gene should have weights built for use in TWAS. Our analysis shows there remains power to be gained by setting this somewhat liberally at 0.1
• --content_alpha Linear model regularization constant for CONTENT. Default is elastic net (0.5), LASSO is 1 and ridge-like is 0.001.
• --tissue_alpha Linear model regularization constant for the context-by-context approach.
• --num_folds Number of folds for cross-validation
• --scale_exp Center and scale the gene expression

Running an example

Download the zip archive containing a sample of input for CONTENT. An example command for CONTENT looks as follows:

Rscript ~/CONTENT/CONTENT_cli.R -X ~/CONTENT/ENSG00000163331.6/genos
-Y ~/CONTENT/ENSG00000163331.6/y/ -o ~/CONTENT/ENSG00000163331.6/ -s
~/CONTENT/ENSG00000163331.6/snps.txt -t ~/CONTENT/TWAS_weights/ -n
ENSG00000163331.6

We tested CONTENT using R (3.6.0) and include analyses with the following software in our manuscript: GCTA (1.93.1), CONTENT (https://github.com/cozygene/CONTENT), UTMOST (https://github.com/yiminghu/CTIMP), bigstatsr (https://github.com/privefl/bigstatsr), FUSION-TWAS (http://gusevlab.org/projects/fusion/), TreeQTL (http://bioinformatics.org/treeqtl/).

[1] Florian Prive, Hugues Aschard, et al. “Efficient analysis of large-scale genome-wide data with two R packages: bigstatsr and bigsnpr.” In:Bioinformatics34.16 (2018), pp. 2781–2787.

[2] C B Peterson, M Bogomolov, et al. “TreeQTL: hierarchical error control for eQTL findings.” In:Bioinformatics 32.16 (2016), pp. 2556–2558.

[3] Michael Wainberg, Nasa Sinnott-Armstrong, et al. “Opportunities and challenges for transcriptome-wide association studies”. In:Nature Genetics 51.4 (2019), pp. 592–599.