A recent trend in large-scale association studies is transcriptome-wide association study (TWAS), which aggregates genomic information into functionally relevant units that map to genes and their expression.
GWAS meets transcriptomics
Genome-wide association studies (GWAS) have been able to identify thousands of genetic variants associated with many complex traits. Nevertheless, for many traits such as neuropsychiatric, identified GWAS-hits have very small effect sizes. In addition, their biological interpretation remains challenging as they are often within non-coding regions, pleiotropic and/or non-causative.
In recent years experts have proposed TWAS as an invaluable tool in investigating the potential mechanisms of these underlying variant-trait associations. Specifically, it is able to integrate GWAS with expression mapping studies to identify gene-trait associations (GTAs).
Opportunities
Gene expression is a heritable, complex trait which varies across different tissues. TWAS is able to predict gene expression for each individual in the GWAS cohort. Statistical associations are then estimated between predicted gene expression and the trait. While these associations do not guarantee causality, TWAS has garnered substantial interest because it prioritises candidate causal genes and tissues underlying GWAS loci.
TWAS aims to alleviate problems of statistical power and interpretability associated with GWAS. Several groups conducting TWAS for breast cancer risk have reported several significant associations for genes with breast cancer susceptibility, which showed increased power over GWAS. Some methods of TWAS, such as UTMOST, increase power by jointly training expression models across multiple tissues.
Challenges and recommendations
Most TWAS studies so far have drawn from data with largely European ancestry. Therefore, it is not clear whether these models can be informative for other groups. While some suggest that stratification by race or ancestry may be necessary to construct proper tests, many of these cohorts do not contain sufficient sample sizes for minority populations. A study published earlier this year in Genome Biology provided a framework for TWAS for breast cancer in diverse populations. The team believe that their work will help inform the utilisation of TWAS methods in polygenic traits and diverse study populations.
A key concern is that correlated expression across individuals may result in false hits. One emerging approach to address co-regulation is repurposing GWAS fine-mapping to TWAS. This is on the basis of the analogy between LD in GWAS and co-regulation in TWAS.
Another vulnerability is bias within expression panels from non-trait related tissues. Commonly, researchers use tissues with large expression panels to maximise power, even if they are less mechanistically related. A paper published last year in Nature Genetics recommended using an expression panel from only the most mechanistically related tissues available (even when they are smaller than other tissues). However, they suggest that if the sample size were to be substantially increased by using a slightly less related tissue, this would be acceptable.
Concluding comments
TWAS is a promising approach in prioritising causal genes at GWAS loci. The development of fine mapping and tackling tissue bias will allow for maximum benefit of this approach. Moreover, developing polygenic risk scoring for TWAS could be a potential avenue in advancing personalised medicine efforts for several disorders.
Image credit: By Photitos2016 – canva.com
