Since their inception, genome-wide association studies (GWAS) have identified more than a hundred thousand single nucleotide polymorphism (SNP) loci that are associated with various complex human diseases or traits. The majority of GWAS discoveries are located in non-coding regions of the human genome and have unknown functions. The valley between non-coding GWAS discoveries and downstream affected genes hinders the investigation of complex disease mechanism and the utilization of human genetics for the improvement of clinical care. Meanwhile, advances in high-throughput sequencing technologies reveal important genomic regulatory roles that non-coding regions play in the transcriptional activities of genes. In this review, we focus on data integrative bioinformatics methods that combine GWAS with functional genomics knowledge to identify genetically regulated genes. We categorize and describe two types of data integrative methods. First, we describe fine-mapping methods. Fine-mapping is an exploratory approach that calibrates likely causal variants underneath GWAS signals. Fine-mapping methods connect GWAS signals to potentially causal genes through statistical methods and/or functional annotations. Second, we discuss gene-prioritization methods. These are hypothesis generating approaches that evaluate whether genetic variants regulate genes via certain genetic regulatory mechanisms to influence complex traits, including colocalization, mendelian randomization, and the transcriptome-wide association study (TWAS). TWAS is a gene-based association approach that investigates associations between genetically regulated gene expression and complex diseases or traits. TWAS has gained popularity over the years due to its ability to reduce multiple testing burden in comparison to other variant-based analytic approaches. Multiple types of TWAS methods have been developed with varied methodological designs and biological hypotheses over the past 5 years. We dive into discussions of how TWAS methods differ in many aspects and the challenges that different TWAS methods face. Overall, TWAS is a powerful tool for identifying complex trait-associated genes. With the advent of single-cell sequencing, chromosome conformation capture, gene editing technologies, and multiplexing reporter assays, we are expecting a more comprehensive understanding of genomic regulation and genetically regulated genes underlying complex human diseases and traits in the future.

Data availabilitySummary statistics generated by COVID-19 Host Genetics Initiative are available online (https://www.covid19hg.org/results/r6/). The analyses described here use the freeze 6 data. The COVID-19 Host Genetics Initiative continues to regularly release new data freezes. Summary statistics for samples from individuals of non-European ancestry are not currently available owing to the small individual sample sizes of these groups, but the results for 23 loci lead variants are reported in Supplementary Table 3. Individual-level data can be requested directly from the authors of the contributing studies, listed in Supplementary Table 1.

Genome-wide association studies (GWAS) have been successful in facilitating the understanding of genetic architecture behind human diseases, but this approach faces many challenges. To identify disease-related loci with modest to weak effect size, GWAS requires very large sample sizes, which can be computational burdensome. In addition, the interpretation of discovered associations remains difficult. PrediXcan was developed to help address these issues. With built in SNP-expression models, PrediXcan is able to predict the expression of genes that are regulated by putative expression quantitative trait loci (eQTLs), and these predicted expression levels can then be used to perform gene-based association studies. This approach reduces the multiple testing burden from millions of variants down to several thousand genes. But most importantly, the identified associations can reveal the genes that are under regulation of eQTLs and consequently involved in disease pathogenesis. In this study, two of the most practical functions of PrediXcan were tested: 1) predicting gene expression, and 2) prioritizing GWAS results. We tested the prediction accuracy of PrediXcan by comparing the predicted and observed gene expression levels, and also looked into some potential influential factors and a filter criterion with the aim of improving PrediXcan performance. As for GWAS prioritization, predicted gene expression levels were used to obtain gene-trait associations, and background regions of significant associations were examined to decrease the likelihood of false positives. Our results showed that 1) PrediXcan predicted gene expression levels accurately for some but not all genes; 2) including more putative eQTLs into prediction did not improve the prediction accuracy; and 3) integrating predicted gene expression levels from the two PrediXcan whole blood models did not eliminate false positives. Still, PrediXcan was able to prioritize GWAS associations that were below the genome-wide significance threshold in GWAS, while retaining GWAS significant results. This study suggests several ways to consider PrediXcan's performance that will be of value to eQTL and complex human disease research.

As a type of relatively new methodology, the transcriptome-wide association study (TWAS) has gained interest due to capacity for gene-level association testing. However, the development of TWAS has outpaced statistical evaluation of TWAS gene prioritization performance. Current TWAS methods vary in underlying biological assumptions about tissue specificity of transcriptional regulatory mechanisms. In a previous study from our group, this may have affected whether TWAS methods better identified associations in single tissues versus multiple tissues. We therefore designed simulation analyses to examine how the interplay between particular TWAS methods and tissue specificity of gene expression affects power and type I error rates for gene prioritization. We found that cross-tissue identification of expression quantitative trait loci (eQTLs) improved TWAS power. Single-tissue TWAS (i.e., PrediXcan) had robust power to identify genes expressed in single tissues, but, often found significant associations in the wrong tissues as well (therefore had high false positive rates). Cross-tissue TWAS (i.e., UTMOST) had overall equal or greater power and controlled type I error rates for genes expressed in multiple tissues. Based on these simulation results, we applied a tissue specificity-aware TWAS (TSA-TWAS) analytic framework to look for gene-based associations with pre-treatment laboratory values from AIDS Clinical Trial Group (ACTG) studies. We replicated several proof-of-concept transcriptionally regulated gene-trait associations, including UGT1A1 (encoding bilirubin uridine diphosphate glucuronosyltransferase enzyme) and total bilirubin levels (p = 3.59×10−12), and CETP (cholesteryl ester transfer protein) with high-density lipoprotein cholesterol (p = 4.49×10−12). We also identified several novel genes associated with metabolic and virologic traits, as well as pleiotropic genes that linked plasma viral load, absolute basophil count, and/or triglyceride levels. By highlighting the advantages of different TWAS methods, our simulation study promotes a tissue specificity-aware TWAS analytic framework that revealed novel aspects of HIV-related traits.

BackgroundMachine learning methods have gained popularity and practicality in identifying linear and non-linear effects of variants associated with complex disease/traits. Detection of epistatic interactions still remains a challenge due to the large number of features and relatively small sample size as input, thus leading to the so-called “short fat data” problem. The efficiency of machine learning methods can be increased by limiting the number of input features. Thus, it is very important to perform variable selection before searching for epistasis. Many methods have been evaluated and proposed to perform feature selection, but no single method works best in all scenarios. We demonstrate this by conducting two separate simulation analyses to evaluate the proposed collective feature selection approach.ResultsThrough our simulation study we propose a collective feature selection approach to select features that are in the “union” of the best performing methods. We explored various parametric, non-parametric, and data mining approaches to perform feature selection. We choose our top performing methods to select the union of the resulting variables based on a user-defined percentage of variants selected from each method to take to downstream analysis. Our simulation analysis shows that non-parametric data mining approaches, such as MDR, may work best under one simulation criteria for the high effect size (penetrance) datasets, while non-parametric methods designed for feature selection, such as Ranger and Gradient boosting, work best under other simulation criteria. Thus, using a collective approach proves to be more beneficial for selecting variables with epistatic effects also in low effect size datasets and different genetic architectures. Following this, we applied our proposed collective feature selection approach to select the top 1% of variables to identify potential interacting variables associated with Body Mass Index (BMI) in ~ 44,000 samples obtained from Geisinger’s MyCode Community Health Initiative (on behalf of DiscovEHR collaboration).ConclusionsIn this study, we were able to show that selecting variables using a collective feature selection approach could help in selecting true positive epistatic variables more frequently than applying any single method for feature selection via simulation studies. We were able to demonstrate the effectiveness of collective feature selection along with a comparison of many methods in our simulation analysis. We also applied our method to identify non-linear networks associated with obesity.