How to Improve Interpretability of Epigenomic Studies
Appropriate study planning from the outset is one step that can help improve interpretability.
Epigenetics is one of the fastest growing research areas in biology. Epigenetic mechanisms control gene expression in a dynamic fashion independently of DNA sequence variation. Abnormal epigenetic regulation has been implicated in several neuropsychiatric disorders such as addiction, anxiety, depression, posttraumatic stress disorder, and schizophrenia.
Because epigenetic processes are heritable and can influence transcriptional regulation, they could potentially disrupt cellular homeostasis and have long-term phenotypic effects. This reasoning has inspired a growing body of research investigating whether epigenetic changes–typically indicated by patterns of DNA methylation and modifications of histones–differentiate individuals with particular phenotypes. This approach is called an epigenome-wide association study (EWAS) and is similar to genome-wide association studies (GWAS) that examine links between phenotypes and genetic variability.
Additionally, the “epigenome is an attractive candidate for mediating long-term responses to cellular stimuli, such as environmental effects modifying disease risk,” according to the authors of a recent review published in PLOS Genetics. However, such studies are part of a broader group of study called “disease-omics,” which has problems that limit the interpretability of the findings. In their paper, the authors from the University of Bristol in the UK and Albert Einstein College of Medicine in New York identified current problems with epigenomic research, and they offer solutions that could enable such studies to fulfill their potential in producing valuable new insights.
In contrast with genetic loci, the vast majority of which remain constant throughout the course of an individual's lifetime, epigenetic patterns may change. “The constancy and random assignment of genetic characteristics allow the case/control study design to succeed, permitting results to be interpreted as causal,” the authors stated. Epigenetic measurements, on the other hand, have similar risks as other phenotypic measurements in a case/control design, such as issues with ascertainment and reverse causation.
In attempting to establish causation between two measurements, the epidemiologist should first eliminate issues common to observational studies: spurious associations–including chance false positives and publication bias, and ascertainment and other selection biases; and reliable but non-causal associations, such as confounding and reverse causation.
One of the other problems with interpreting EWAS results, according to the paper, pertains to the impact of cell subtype. “One major focus has been on the potential for cell subtype proportional heterogeneity to influence the DNA methylation patterns observed in pools of cells,” as supported by numerous recent studies, the authors reported. They found evidence of cell subtypes with distinct patterns of methylation, and it is “likely that even when using purification techniques, a pool of cells is composed of multiple epigenomes, generating what we refer to as a ‘meta-epigenome,'” they said.
Even carefully designed studies usually detect only modest methylation changes, and findings indicate a mosaic cellular response for phenotype-associated epigenetic changes. Limited degrees of methylation change can also result from–rather than cause–transcriptional changes. Even more concerning is the influence of DNA polymorphism, which accounts for approximately 22% to 80% of the methylation variability between people. The authors say that no EWAS study to date can be considered fully interpretable, as the findings may be due to effects of cell subtypes or transcriptomic or genomic variability.
Appropriate study planning from the outset is one step that can help improve interpretability. The cross-sectional case/control approach is not generally a strong choice because of ascertainment issues and the possibility of reverse causation, and these problems cannot be fixed with a larger sample size or different type of cell. A longitudinal study design can provide insights relevant to both biomarkers and mechanisms, and it allows researchers to identify epigenetic changes that occur before phenotype development.
The authors offer a checklist incorporating these various elements on how to improve the interpretability of EWAS results, including the advice to “perform transcriptomic studies on the same cells tested for epigenetic changes and genotyping of the same individuals,” which “allows a number of causes and consequences of changes of epigenetic regulators to be interpreted."
Birney E, Smith GD, Greally JM. Epigenome-wide association studies and the interpretation of disease –omics. PLOS Genet. 2016; 12(6): e1006105.