Recommendations to encourage participation of individuals from diverse backgrounds in psychiatric genetic studies.

We present innovative research practices in psychiatric genetic studies to ensure representation of individuals from diverse ancestry, sex assigned at birth, gender identity, age, body shape and size, and socioeconomic backgrounds. Due to histories of inappropriate and harmful practices against marginalized groups in both psychiatry and genetics, people of certain identities may be hesitant to participate in research studies. Yet their participation is essential to ensure diverse representation, as it is incorrect to assume that the same genetic and environmental factors influence the risk for various psychiatric disorders across all demographic groups. We present approaches developed as part of the Eating Disorders Genetics Initiative (EDGI), a study that required tailored approaches to recruit diverse populations across many countries. Considerations include research priorities and design, recruitment and study branding, transparency, and community investment and ownership. Ensuring representation in participants is costly and funders need to provide adequate support to achieve diversity in recruitment in prime awards, not just as supplemental afterthoughts. The need for diverse samples in genetic studies is critical to minimize the risk of perpetuating health disparities in psychiatry and other health research. Although the EDGI strategies were designed specifically to attract and enroll individuals with eating disorders, our approach is broadly applicable across psychiatry and other fields.

A method to improve the reproducibility of findings from epigenome- and transcriptome-wide association studies.

Reproducibility is a cornerstone of scientific progress. In epigenome- and transcriptome-wide association studies (E/TWAS) failure to reproduce may be the result of false discoveries. Whereas multiple methods exist to control false discoveries due to sampling error, minimizing false discoveries due to outliers and other data artefacts remains challenging. We propose a robust E/TWAS approach that outperforms alternative methods to improve reproducibility such as split-half replication. Furthermore, robust E/TWAS results in only a minor loss of power if there are no outliers and can in the presence of outliers, likely a more realistic scenario, even be more powerful than regular E/TWAS.