--- title: "INTACT: Integrate TWAS and Colocalization Analysis for Gene Set Enrichment Analysis" shorttitle: "INTACT" author: - name: Jeffrey Okamoto affiliation: - &id Department of Biostatistics, University of Michigan email: jokamoto@umich.edu - name: Xiaoquan Wen email: xwen@umich.edu affiliation: *id abstract: > This package integrates colocalization probabilities from colocalization analysis with transcriptome-wide association study (TWAS) scan summary statistics to implicate genes that may be biologically relevant to a complex trait. Given gene set annotations, this package can estimate gene set enrichment using posterior probabilities from the TWAS-colocalization integration step. package: INTACT output: BiocStyle::html_document vignette: > %\VignetteIndexEntry{INTACT: Integrate TWAS and Colocalization Analysis for Gene Set Enrichment} %\VignetteEngine{knitr::rmarkdown} %\VignetteEncoding{UTF-8} toc_float: true --- # Installation To install this package, run the following code chunk (in R 4.2 or later): ```{r,eval=FALSE} if (!require("BiocManager", quietly = TRUE)) install.packages("BiocManager") BiocManager::install("INTACT") ``` # Methodology Reference For a comprehensive description of the probabilistic framework behind INTACT please refer to: Okamoto, Jeffrey, et al. "Probabilistic integration of transcriptome-wide association studies and colocalization analysis identifies key molecular pathways of complex traits." *The American Journal of Human Genetics* 110.1 (2023): 44-57. # Introduction Integrative genetic association methods have shown great promise in post-GWAS (genome-wide association study) analyses, in which one of the most challenging tasks is identifying putative causal genes and uncovering molecular mechanisms of complex traits. Prevailing computational approaches include transcriptome-wide association studies (TWASs) and colocalization analysis. TWASs aim to assess the correlation between predicted gene expression of a target gene and a GWAS trait. Common output for TWASs include gene-level z-statistics. Colocalization analysis attempts to determine whether genetic variants that are causal for a molecular phenotype (such as gene expression) overlap with variants that are causal for a GWAS trait. Common output for colocalization analysis often include gene-level colocalization probabilities, which provide evidence regarding whether there exists a colocalized variant for the expression of a target gene and GWAS trait. Recent studies suggest that TWASs and colocalization analysis are individually imperfect, but their joint usage can yield robust and powerful inference results. INTACT is a computational framework to integrate probabilistic evidence from these distinct types of analyses and implicate putative causal genes. This procedure is flexible and can work with a wide range of existing integrative analysis approaches. It has the unique ability to quantify the uncertainty of implicated genes, enabling rigorous control of false-positive discoveries. INTACT-GSE is an efficient algorithm for gene set enrichment analysis based on the integrated probabilistic evidence. This package is intended for performing integrative genetic association analyses in tandem with other Bioconductor packages such as `biomaRt` or `GO.db`, which could be used to obtain gene set annotations for gene set enrichment analysis. # Included Data Sets To illustrate the functionality of the `INTACT` package, we include a simulated data set `simdat`. See the methodology reference for an explanation of the simulation design. The data is organized as a 1197 row by 3 column data frame, where rows correspond to genes, the GLCP column provides gene-level colocalization probabilities, and the TWAS_z column provides TWAS scan z-scores. Additionally, we include a simulated gene set list `gene_set_list`, which contains two gene sets. The first gene set has 503 gene members and is significantly enriched among the genes included in `simdat`, based on the probabilistic INTACT output. The second gene set has 200 gene members and is not enriched among the `simdat` genes. We include `gene_set_list` to show how to perform gene set enrichment estimation using INTACT-GSE. # INTACT: Integrating TWAS Scan and Colocalization Analysis Results The first main functionality of this package is integrating results from a transcriptome-wide association study (TWAS) scan and a colocalization analysis. The TWAS scan results must be in the form of gene-level z scores, and the colocalization analysis results should be in the form of gene-level colocalization probabilities. These are provided as output by most popular TWAS and colocalization methods. Below, we include an example of how to use INTACT to integrate TWAS scan and colocalization results for a simulated data set `simdat`. ```{r} library(INTACT) data(simdat) rst <- INTACT::intact(GLCP_vec=simdat$GLCP, z_vec=simdat$TWAS_z) ``` The `intact` function takes a vector of gene-level colocalization probabilities `GLCP_vec` and TWAS scan z-scores `z_vec`. It outputs gene-level posterior probabilities of putative causality. The example included above uses default settings for the prior function and truncation threshold $t$ (`prior_fun = linear` and `t=0.05`). There are three additional prior functions implemented in the `INTACT` software, including `expit`, `step`, and `hybrid`. The `expit` and `hybrid` options have an additional curvature shrinkage parameter `D`, with a default value of 0.1. The default truncation parameter value for the step prior function is 0.5, while the default value is 0.05 for all other prior functions. Below are three additional examples of how to integrate the TWAS z scores and colocalization probabilies from the simulated data using different prior function, truncation threshold, and curvature shrinkage settings: ```{r} rst1 <- INTACT::intact(GLCP_vec=simdat$GLCP, prior_fun=INTACT::expit, z_vec = simdat$TWAS_z, t = 0.02,D = 0.09) rst2 <- INTACT::intact(GLCP_vec=simdat$GLCP, prior_fun=INTACT::step, z_vec = simdat$TWAS_z, t = 0.49) rst3 <- INTACT::intact(GLCP_vec=simdat$GLCP, prior_fun=INTACT::hybrid, z_vec = simdat$TWAS_z, t = 0.49,D = 0.05) ``` If the user wishes to specify TWAS Bayes factors instead of z-scores, they can do so through the argument `twas_BFs`. The Bayes factors should be a numeric vector with genes in the same order as the colocalization probabilities vector. If the user wishes to specify gene-specific TWAS priors, they can do so through the argument `twas_priors`. If no input is supplied, INTACT computes a scalar prior using the TWAS data (see the methodology reference for more details). We provide an additional function `fdr_rst` that is useful if the user wishes to perform Bayesian FDR control on the INTACT output. An example of how to apply this function at a target control level of 0.05 is shown below. ```{r} fdr_example <- fdr_rst(rst1, alpha = 0.05) head(fdr_example) ``` # INTACT-GSE: Gene Set Enrichment Estimation Using INTACT results The `INTACT` package provides the `intactGSE` function to perform gene set enrichment estimation and inference using integrated TWAS scan z-scores and colocalization probabilities. This function requires a data frame `gene_data` containing gene names and corresponding colocalization probabilities and TWAS z-scores for each gene. Column names should be "gene", "GLCP", and "TWAS_z'. If the user wishes to specify TWAS Bayes factors instead of z-scores, use the column name "TWAS_BFs". If the user wishes to specify gene-specific TWAS priors, use the column name "TWAS_priors". In addition to `gene_data`, the user must provide a list of gene sets `gene_sets`. The format of `gene_sets` must match the included example `gene_set_list`: it must named list of gene sets for which enrichment is to be estimated. List items should be character vectors of gene IDs. Gene ID format should match the gene column in `gene_data`. The user can specify the same prior-related arguments as in the `intact` function, including `prior_fun`, `t`, and `D`(only when the prior function is specified as `expit` or `hybrid`). The user can specify the method by which the standard error of the enrichment estimate is computed. Options include a numerical differentiation of the score function (default): `NDS`; a profile likelihood approach: `profile_likelihood`, and bootstrapping: `bootstrap`. For hypothesis testing, the user can specify a significance threshold, which is 0.05 by default. An example of how to estimate gene set enrichment in the gene sets provided in `gene_set_list` (using default settings) is shown below: ```{r} data(gene_set_list) INTACT::intactGSE(gene_data = simdat,gene_sets = gene_set_list) ``` The output of `intactGSE` includes one row per gene set and eight columns: the gene set name, the enrichment parameter $\alpha_1$ estimate, the enrichment parameter estimate standard error, the z-score, the p-value, the left and right CIs, and the convergence flag (if CONVERGED = 1, then the enrichment estimation algorithm converged. If not, CONVERGED = 0). Some data sets are not informative for gene set enrichment estimation; in this case, the algorithm will fail to converge. We emphasize that failure of the algorithm to converge does not provide information regarding the enrichment (or lack thereof) for a given gene set. Finally, we include three additional examples of how to estimate enrichment for the same data sets using non-default prior parameters: ```{r} INTACT::intactGSE(gene_data = simdat,prior_fun = INTACT::step, t = 0.45,gene_sets = gene_set_list) INTACT::intactGSE(gene_data = simdat,prior_fun = INTACT::expit, t = 0.08,D = 0.08, gene_sets = gene_set_list) INTACT::intactGSE(gene_data = simdat,prior_fun = INTACT::hybrid, t = 0.08,D = 0.08, gene_sets = gene_set_list) ``` Session information is included below: ```{r} sessionInfo() ```