Brian Schilder is a first year PhD student in the lab of Dr Nathan Skene. His research title is "Multi-omic medicine: cell-type-specific mechanisms of neurodegenerative disease genomics".
July 2021
Dr Abbas Dehghan is a UK DRI co-investigator and a reader in Cardiometabolic Disease Epidemiology. He specialises in linking genetic, epidemiology and metabolic phenotyping in dementia in the context of ageing, environment and lifestyle.
Brian: Where did you start, did you always know you wanted to go into this field, or have you moved around in your career?
Abbas: When I started in medicine, one of the topics that got my attention was epidemiology. I got the sense that you could approach diseases on a large scale rather than on a case-by-case basis. That made a lot of sense to me and encouraged me to read more about epidemiology research. When I graduated, I knew that this was the direction I wanted to go.
I did pure epidemiology, and then later, when I started my PhD at Erasmus University in the Netherlands, our department received funding to do genome-wide association studies (GWAS)(and the Rotterdam Study), and I modified my PhD project to some extent to do genetics, because it was a fascinating project. We had a lot of hope with GWAS back in 2005-06.
Brian: How far were you into your PhD when you decided to make that shift?
Abbas: About a year. I had been doing my PhD for a year and the main focus was meant to be novel risk factors for cardiovascular disease and Type-II diabetes. I had looked at a couple of markers and we also brought in genetic factors. The Rotterdam Study was a founding member of the CHARGE consortium (The Cohorts for Heart and Aging Research in Genomic Epidemiology), which made trans-Atlantic collaborative projects possible with lots of excellent scientists from leading centers. People whose papers I was reading 2 years before and then was working with; that was very exciting.
After my PhD I knew that I wanted to continue in that field, not just with genetics but with a wider range of molecular epidemiology, so I started to work on DNA methylation and cardio-metabolic traits in the cardiovascular group of The Rotterdam Study. In 2016 I moved to Imperial. I was working with Professor Paul Elliot from a few years before on a European-funded project called COMBI-BIO, where we were looking at metabolomics in relation to atherosclerosis. I thought that metabolomics was the next step in molecular epidemiology that I want to learn and work on.
Brian: What drew you to metabolomics specifically, as opposed to some other layer of biology?
Abbas: If you look at the layers of information at the cellular level that we can study; you start with genetics, epigenetics, transcriptomics, proteomics, and then at the end you come to metabolomics. As you move down, there is less stability in the data but there is also more information. When it comes to metabolomics, you have a lot of environmental exposure information mixed with genetic information, and that is the closest step to health and disease and all phenotypes we are studying. I think that is a very rich set of data, that of course comes with complexities, but I thought there were a lot of opportunities in the metabolomics data, both in etiologic research and risk prediction. We had tried to use genetic information for instance, to improve cardiovascular disease risk prediction before, but it didn’t improve it in a clinically meaningful way. I think one of the reasons is that we were ignoring the lifestyle and environmental factors that are important. With metabolomics, we have this We’re still not there, but I think that it might be an interesting place to look.
Brian: It’s a very complex problem.
Abbas: Yeah, and when I moved to Imperial I wasn’t limited to working only on cardiovascular disease and extended my work to other complex diseases such as dementia. For instance, I now work with the UK Dementia Research Institute (UK DRI). I think that’s an excellent opportunity. We are the only epidemiology-based program in the institute. Other groups are doing basic science and experimental work, which provides lots of opportunities for collaborative work.
Brian: You started off getting your medical degree and then moved onto getting a PhD; it wasn’t a joint program right?
Abbas: No, after I finished medicine, I had one or two years that I was working in clinic within the public health system, and eventually I made up my mind that I wanted to do research. I think I’m better working with numbers than with patients. And that was a big decision for me, because to be honest working in a clinic is very rewarding in a short time. You do something for a patient, you see the results and that’s very satisfying. With research, you start a project now and a paper might be published in a few years, so you must keep the momentum and keep working. Even then, you don’t know how many people will read it and what the impact of the paper will be. But you have to trust science, and do your best and be confident that it will have impact at the end of the day.
Brian: And potentially have an impact on more people than you as a physician could reach individually.
Abbas: Exactly, that was the first thought I had. That’s why I like epidemiology.
Brian: In your opinion, what do omics (e.g. genomics) offer that epidemiology alone doesn’t, and how do these fields complement each other?
Abbas: I believe that omics technologies in general bring more precision to the measurements that could be done in epidemiologic studies. So it’s like giving a magnifier to the epidemiologist, to study a particular exposure and risk factor in detail. At the same time you are increasing the scope of your research and adding to the precision and the details that you can bring into your research.
Brian: And how about from the other end, what aspects of epidemiology (either in the methodology or the way of thinking) offer things that are not necessarily the primary focus within omics?
Abbas: I believe what is important in applying omics technologies to large population-based studies, is that you have the chance to study diseases within the community that you’re interested in. In an experimental work on several samples, you might be able to, for instance, find a mechanism for a disease. But you never know how many people get the disease through that mechanism. It could be what we describe as a “storm in a glass of water”. But in epidemiology, the data points you to the pathways or risk factors that are most important for the population, so you will get to the right risk factors or mechanisms that should be prioritized for any kind of public health or prevention/treatment. I think that’s the added value of epidemiology studies that you cannot get from working on a limited number of samples in your lab.
Brian: What are some of the most interesting or surprising findings that you’ve come across in your research? And then the follow up to that is, what are the most interesting or surprising relationships between phenotypes or diseases in the field in general?
Abbas: Well I should say, that in most of the research we do most of the findings are really exciting. One that I can highlight here is a project that we did on education and cardiovascular disease. It’s known that high education is associated with lower risk of cardiovascular disease. We were interested to know what the mediating factors are; why do we see less cardiovascular disease in one population just based on education? We looked at known risk factors (such as smoking, BMI, blood pressure), used a novel method called Mendelian Randomization and calculated the mediation effect through these risk factors. We discovered that a large part of the effect doesn’t go through the known risk factors. It means that there is more that we don’t know, with regards to cardiovascular disease. I’m interested to apply that to other diseases as well. For instance, we could apply it to dementia, because education level is also a risk factor for Alzheimer’s disease and dementia. We don’t know the risk factors that are mediating these effects, but I think it would be interesting to test this way.
Brian: I’m curious, because when we do these kinds of studies looking at relationships between phenotypes, the causality relationships are still quite difficult to disentangle. And I’m wondering, how you approach that. For example, is education a proxy for something, or is it reflective of a tendency to seek more education. What are the relationships?
Abbas: I think the right answer is, we don’t know. Mendelian Randomization is an excellent approach that has been developed in recent years to study causality. But when it comes to complex traits, such as education, the precision and the definition of the exposure is always challenge in interpreting the findings. So as you pointed out, what aspect of education is important here?
What we tried to address in that project was the next step. In the case of education, is it intelligence? Is it social group? What’s the next step it takes to reduce the risk of cardiovascular disease? Does that for instance lead to a healthier lifestyle (less smoking, better diet)? Because that step could be further encouraged in other groups to prevent the disease.
Brian: So you’re very much focused on the interventional side; understanding the mechanisms in order to modify someone’s risk for a disease?
Abbas: Yes, well the other part is also interesting, but the implications of the mediating factors is more important. Maybe that’s because I’m an MD.
Brian: I think it’s a good grounding to have with any basic research. It’s possible to get in the weeds and forget why we’re doing this to begin with because we’re so focused in our particular area. And I think having that grounding in the clinic and thinking about the patient is really good to have as a driving force.
Abbas: Yes exactly.
Brian: Are there any areas in your fields of expertise (e.g. epidemiology, omics, medicine) that you feel the community is heading the wrong direction? In other words, perhaps being a bit too focused on a particular direction. Then the follow up to that is, where do you think the community is heading the right direction, where there is a lot of potential for impact?
Abbas: Something that comes to mind, is that the field is a bit technology driven. We need to be careful why we apply each method; is there a good scientific reason to move in that direction, or is it just happening because the technology is at hand? I don’t have any particular examples where I can say, “this has gone wrong”, but I think we need to be careful to take steps in a very evidence-based way.
Brian: Are you thinking more on the data acquisition side in terms of technologies, or the software side?
Abbas: I was more thinking of the data acquisition, because that’s a massive part. These are investments that are made, and we need to make sure that the investment is moving us in the direction that matches with our priorities and what science is pointing us towards. With more technologies and more methods becoming available you also need to prioritize where we need to invest.
Another direction that I can highlight here is, the investment has been done to increase sample size. 20 years ago, epidemiologic studies were all around 3-5,000 to 10,000 individuals, and that was large enough for the type of research we were doing. Now we’re dealing with hundreds of thousands, with databases like UK Biobank. Every country is trying to have their own biobank somehow. These are a lot of investments, and they’re very valuable especially in terms of omics studies where we’re dealing with small effects. Small but meaningful relationships. That should be done, I have no opposition to that, but we shouldn’t forget about smaller more focused studies that are aiming to tackle a relevant research question. That’s one corner that shouldn’t be overlooked. If we need to phenotype a population in a better way, on the scale of let’s say 1-2,000 individuals, with more robust data (e.g. insulin resistance, imaging) it’s still not possible to apply this to tens or hundreds of thousands, or particular patient groups.
Brian: How about on the phenotype side? Are there certain things that you think should be collected more often in these databases?
Abbas: Step by step we are collecting better data. For instance, physical activity or sleep. Before we were collecting data based on surveys, and nowadays it much easier to collect objective data using accelerometers. It also makes it much easier to identify different kinds of physical activity, or the timing of the physical activity.
Brian: So you’re talking about peripheral devices, something like a smart watch?
Abbas: Yeah, something like a smart watch, depending on what kind of activities you want to collect. But let’s say the participants put it on their wrist and then carry it for a week. You collect the data for that period of time, and you will have exact estimates of sleep, or time spent sitting in front of the TV, sedentary lifestyle, and activities such as running, biking etc. There’s also the possibility to use datasets that come from their smart phones or smart watches. We are now in an era that we have more data than analysts. Twenty years ago, we were looking around for data to analyze. Now everyone has the data and they’re looking for someone to analyze it and make sense of it.
Brian: An increasingly challenging problem. But also, increasingly exciting.
Abbas: Yes, exactly.
Brian: Given your experience in different fields, both on the medical side and the research side, do you have any parting words of wisdom for young or early career researchers, who are trying to figure out where they can apply their interests in a way that is going to have impact?
Abbas: Well, I think perhaps I’m too young myself to give advice to younger researchers, but what I’ve done and what I can advise others to do, is to follow your passion. Do something you love to do. Then it doesn’t feel like working, it feels like fun. You never get tired of it, and every day brings a lot of excitement.
Brian: I think that’s excellent advice.