Psychiatric Genomic Consortium Workshop Summary
State of the Science: Exploring Best Practices in Traumatic Stress Genomics – June 20, 2017 – Cambridge, MA
The Psychiatric Genomic Consortium Workshop on “State of the Science: Exploring Best Practices in Traumatic Stress Genomics” was conducted by Cohen Veterans Bioscience and the Stanley Center for Psychiatric Genetics at the Broad Institute. The goal of the meeting was to explore the state of the art in PTSD genetics and discuss best practices in genomics research. A series of presentations were made by thought leaders, culminating in engaging discussions regarding best practices in PTSD genomics.
The objectives of this one-day meeting included:
- Review PTSD research challenges in the age of consortium science
- Examine the state of genetics research across other complex diseases
- Discuss best practices in the design and analysis of genetic studies
A consistent theme emerged throughout the day regarding the need for open, large-scale and collaborative science, with rigorous statistical approaches to be successful in advancing the field of PTSD and improving patient outcomes. Highlights of the workshop are summarized here.
State of Science: Genetics as the foundation for translational research in psychiatry, Dr. Steve Hyman
- The need to create a new roadmap from genes to therapeutics for PTSD was discussed. Some of the challenges to overcome are to understand disease mechanisms of cognition and behavior and the underlying genetic basis of disease using unbiased large-scale non-reductionist genetic studies such as genome wide association studies (GWAS). Large-scale analysis of complex trait genetics requires large sample sizes, open data sharing, rigorous statistics and collaboration.
Complex Trait Genetics: What do we know now?, Dr. Mark Daly
- Genome-wide association studies in PTSD compare genetic differences between thousands of cases with and without diseases to detect actionable genetic variants that show a causal relationship to disease. An integrated genetic approach is required to weave together inherited genetic risk, polygenic variants, quantitative traits and rare and spontaneous variants to identify which individuals are disposed to different disease trajectories.
The State of PTSD Genetics Research, Dr. Caroline Nievergelt
- The Freeze 2 study results and next steps were highlighted. The Freeze 2 study is designed to accelerate the discovery of actionable PTSD biomarkers from 22,000 cases and 59,000 controls worldwide. Among 13,000 European cases, a zinc finger protein of unknown function was identified that has also been implicated in bipolar disorder. Correlation analysis between polygenic risk score and 229 traits revealed age of first birth as well as mental health measures such as subjective well-being and depressive symptoms were highly correlated with PTSD.
- Moving forward, additional analyses are planned to careful profile subjects to exclude controls without trauma-exposure (who may carry a high genetic risk for PTSD, but never exhibit symptoms in the absence of trauma) and perform GWAS on symptom clusters of PTSD.
- An important best practice is to pre-specify study characteristics to prevent replication shopping for patient data that could introduce bias of not reporting negative results.
Best Practices in Genetics, Dr. Ben Neale
- GWAS studies involve multiple testing with many common variants and pose a statistical challenge. Best practices of large scale studies require discipline and rigor about analyses being conducted to distinguish the true effects of complex traits from spurious results.
- Rigorous statistical approaches are essential to mitigate the effects of multiple testing, failure to control for confounders, and problems of negative replication. There are many sources of bias including population stratification; technical bias; source of DNA, heterozygosity, non-random missingness, reproducibility of genotypes, Hardy-Weinberg Equilibrium deviation, genotype calls, sequencing artifacts, etc.
PTSD Phenotype Challenges, Dr. Arieh Shalev
- PTSD is a complex disease characterized by multi-causality, equi-finality (ie., many paths to get PTSD) and multi-finality (ie., diverse outcomes of PTSD, MDD, etc), which lead to unreliable risk indications, heterogeneity, and inconsistent interventions.
- Data-driven, machine learning tools can be developed to predict which patients are likely to experience PTSD symptoms and those who are likely to be resilient. Dr. Shalev presented results from a predictive model that combined early symptom severity and baseline CAP scores to predict the likelihood/probability of developing PTSD within 6 months of an event.
Best Practices in GxE, Dr. Peter Kraft
- When studying genes and environment, it is important to consider biological, public health and statistical factors. Aggregating risk variants can increase power to detect gene by environment interactions. Specific recommendations were advanced to improve risk prediction and prognostic models: large sample sizes with harmonized phenotypes and exposures; well-powered, well-designed, focused replication studies; significance thresholds to account for multiple testing; patterns consistent with hypothesized mechanisms; focus on genetic variants with known marginal effects; appropriate statistical test for discovery or characterization; report effect sizes including main gene and environment effects; publish null findings.
Using phenotypes to inform genetic studies, Dr. Jake Taylor
- The development of efficient and inexpensive standardized instruments was described to improve evidence-base phenotyping using empirically-established molecular evidence for large scale psychiatric studies.
- Next steps are to develop instruments for assessing symptom level and course-of-illness data, trauma load, cognitive, psychological and behavioral responses to trauma in GWAS studies.
Panel Discussion with Drs. Neale, Kraft, Shalev, Taylor, Nievergelt and Koenen (moderator)
- The panel discussed next steps for analysis of the Freeze 2 dataset. It will be important to investigate how measured trauma exposure is a liability to presentation of PTSD. Another approach is to assess symptom cluster, disease course, or subgroups to explore polygenic risk with other conditions or traits in the general population. To assess gene by environment covariation, twin studies in this dataset may also be specifically analyzed to elucidate this question. The role of comorbidities and gender and type of trauma (personal vs impersonal) on PTSD should be explored.
- Consortium-based research presents opportunities for junior researchers to access the data, generate hypotheses, get involved in different committees, community service, or infrastructure to harmonize data. The PTSD working group reviews secondary analysis proposals on a regular basis.
Funders Perspective: Roundtable Discussion, Drs. Haas, Smyth, Hyman and Koenen (moderator)
- The roundtable discussed the three-way, flexible collaboration between the Stanley Center, the Veterans Administration, and Cohen Veterans Bioscience to address operational impediments that are inherent to large scale studies of PTSD. Ongoing partnerships will be critical for the advancement of our understanding of the biological underpinnings of PTSD and development of diagnostics and novel interventions.
If you are interested in more details, please contact firstname.lastname@example.org.