- Course: iBSc Genomic Medicine
- Unit: Genomic Data Science
- Authors: Luisa Zuccolo, Kaitlin Wade, Gibran Hemani, Paul Yousefi, Tom Gaunt
- Contact details: g.hemani@bristol.ac.uk
The issue that we identified
The Genomic Data Science unit teaches students programming and statistical methods for population-scale genetic data. There is a long-standing problem in this field that genetic studies in the vast majority focus on European individuals (https://gwasdiversitymonitor.com/). This is partly due to technical reasons – restricting to a single ancestral group avoids certain statistical problems, but likely also reflects a racial bias that exist more widely in the medical field. Our course material reflected and propagated this bias. The data that we used throughout the course was of European ancestry and, for simplicity, there was no teaching material that described the statistical approaches that would enable our students to handle more diverse genetic data.
The consequences of genetic studies focusing solely on European samples are potentially serious for two reasons. First findings in Europeans may not generalise to more diverse populations, and as a consequence any medical benefits from this work will be restricted to a single ancestral group. Second, more recent statistical techniques that enable the inclusion of diverse samples actually solves a number of problems that can’t be solved when restricting to homogeneous samples.
Our students are the future leaders of this field. We felt it important that they understand the technical reasons for limiting analysis to a single ancestral group, but also learn that including multiple ancestries is not only ethically important but scientifically imperative also.
Actions
We designed a session that asked students to analyse both the merits and pitfalls of restricting genetic analysis to one ancestral group from a purely statistical / technical perspective. Students were asked to form two groups, each tasked with preparing a presentation. The first group would present the statistical benefits of restricting analysis to a single ancestral group; the second group would present the statistical benefits of including multiple ancestral groups in the analysis.
Impacts
We received very positive feedback from the students at the end of the course, in particular for this session. Students appreciated the opportunity to contextualise the technical (and often quite dry) statistical subject matter against the backdrop of social and medical consequences.
What we have learned
During the presentations the students chose to focus more on the sociological impacts than the more technical analytical questions. As a consequence there is a concern that while they appreciate the importance of including diverse samples, they have not necessarily acquired the skills to do this in practice.
A common pitfall relating to decolonisation activities is to simply add a token session at the end of the course that covers the question, which is essentially what has been done here. There does exist an obstacle for us to embed the analysis of diverse samples throughout the course, in that it is substantially more complex than their current materials, and the course is already seen as a substantially complex course. Yet we risk teaching our students out-dated methods by not embracing this added complexity and normalising diversity. This experience has revealed to us that embedding diversity at the core of teaching is important from both an ethical and scientific approach.