Polygenic Prediction of Obesity Across the Life Course

Key Takeaways

Polygenic Prediction and the Biology of Obesity
Obesity is increasingly understood as a disease of the brain, where genetic differences influence hunger, satiety, and reward responses to food. Rather than acting through adipose tissue, many obesity-associated genes regulate hedonic food intake and energy balance through central neural pathways.

Genetic Architecture of Adiposity
Genome-wide association studies (GWAS) have identified over 9,000 genetic variants associated with adiposity. A comprehensive polygenic score based on 5.6 million variants captures the cumulative genetic predisposition to obesity. While heritability reflects population-level variation, polygenic scores quantify individual-level risk.

Predictive Performance
In people of European ancestry, current polygenic scores explain up to 17.5% of the variance in BMI – a major advance in understanding the genetic contribution to obesity. However, the score reflects probability, not destiny; environmental and behavioural factors remain powerful modifiers of genetic risk.

Pathophysiological Insights
These findings reinforce that obesity is not solely driven by willpower or lifestyle but by biological mechanisms influencing appetite control and reward processing. Understanding these mechanisms can help reframe obesity as a chronic, neurobiological disease and support more personalised prevention and treatment strategies.

Future Directions and Next Steps

• Integrate polygenic scores into obesity risk assessment and early prevention programmes
• Investigate how polygenic risk interacts with environmental, behavioural, and pharmacological interventions
• Extend studies to diverse ancestral groups to improve generalisability and equity in prediction tools
• Advance communication strategies to convey that genetic risk does not equal inevitability, reducing stigma and supporting person-centred care
• Strengthen interdisciplinary collaboration to translate genetic insights into practical clinical and public health applications

Summaries are AI-generated from meeting transcripts.

Transcripts are auto generated, if you spot an error, please email enquiries@easo.org

Speaker 1
00:00 – 01:16
Okay, so good afternoon everyone. Welcome to this month’s EASO Early Career Network eLearning Hub event and thank you for joining. Today we have Professor Ruth Luce who will be delivering what will be a fantastic session on polygenic prediction of obesity across the life course. The Nervo Nordisk Foundation has provided support to EASO for these developmental activities including this webinar series and we’re very thankful for this. So my name is Matt Harris. I’m a clinical researcher and one of the board members of the Early Career Network. I’m also joined by fellow board member Beatrice Farina who will introduce herself in a second. Today’s webinar will be recorded and will be available afterwards along with the other eLearning Hub online events and if you haven’t already please do consider joining the ECN which is free along with these excellent webinars. So this will be informal and please do use the Q&A function for questions for Professor Luce as we go throughout the session. There’ll be an opportunity to ask these questions towards the end and also please do fill in the feedback form. These comments are really valued and support the development of future webinars. So it’s my pleasure to hand over to Beatriz who will give us an update of the EASO Early Career Network opportunities.

Speaker 2
01:18 – 04:57
Hello everyone, as Matthew very kindly introduced me I’m Beatriz Farinha, so I’m a PhD student from Spain, also on the board member of the Early Career Network, Alvarezo, and I’m here to share a few updates and opportunities for all the ECN members. So first we have several, we do have several exciting opportunities available and a link will further be on the chat for so everyone can then check them out also much in more detail. So the ECN board is very much welcome, looking forward to welcoming all the people, all the ECN members that are coming to our ECN master class on hot topics in obesity, which will take place in Portugal from now in next week in November. So everyone that goes to Caxcais, have a really nice time, it’s a really nice city, enjoy the weather and I hope it’s not raining, but at least it will be warm. If you haven’t had the chance to participate in one of these events, we really encourage you to apply in the future. It’s a really great learning and networking experience, so please go forward. Next, also, like we announced last, the last webinar, the abstract submissions for the European Congress on Obesity are already open until 11 January 2026. The ECHO 2026 will be in Istanbul, Turkey from the 12th to the 15th of May. We’re very excited to see you all there. Also the ECN Best Thesis Award is also now open for submissions and if you’re nearing completing of your thesis or have recently completed it, please consider applying it by 11 January and the top three candidates will be invited to present their work at a dedicated session at the ECHO 2026 in Istanbul. and the winner will also receive a 500 euros cash prize, which is also really nice. So we are also pleased to also share that we are going to continue with Novo Nordiski Foundation with the prize, with the new investigator awards until 2030. So this award covers four categories, basic science, clinical research, childhood obesity and public health and each come with the research grant. So, and all the winners will be also supported to attend and speak at the special award session at ECHO 2026 in Istanbul. For more ECN activities, we also have the ECN spotlight feature, which highlights ECN members and their research. This is shared across all EASO websites and social media. And so there is a really good chance that you can see other people’s other early career researchers or people that work with obesity in this type of feature. And we would like to have your work also featured. So please be in touch with us if you want. And also if you know someone who is interested. And finally, we also have an ECN WhatsApp group for quick updates, sharing resources and connecting with other members across Europe. So we would love for you to join. And thank you all very much for your attention and we look forward to seeing many of you in our upcoming events. And now I pass to Matt.

Speaker 1
04:58 – 05:16
That’s great. Thank you, Beatrice. Lots and lots of very exciting opportunities. So without further ado, it’s my pleasure to introduce our speaker, Professor Ruth Loos, who is the Vice Executive Director and Group Leader at the Novo Nordisk Foundation Centre for Basic Metabolic Research at the University of Copenhagen. Thank you very much.

Speaker 3
05:18 – 53:02
Thank you so much for having me. I first share my slides. Okay, I think that should work. If not, just give me a shot. So my name is Ruud Loos, and it’s a real pleasure to be here and to be able to present our most recent work. I’ll be focusing on the recent paper that we published in Nature Medicine, together with another paper that we also published in Nature Medicine. So the title of my talk is Polygenic Prediction of Obesity Across the Life Force. These are my disclosures. of these affect my research. And just as a brief introduction, I’m a geneticist, I’m a genetic epidemiologist and study the causes of obesity. But I of course, obesity is not all genetics. And I of course, also believe that the environment, the changing environment is a major driver behind the epidemic epidemic where we are in right now. So like over the the last 40, 50 years, the prevalence of obesity globally has doubled. In no doubt, that is due to the changing environment. But as I said, it’s not all environment. It’s particularly people who are genetically predisposed, who will gain the most of weight in this obesogenic environment. We know that from twin studies, from family studies, and even also from adoption studies that the heritability of obesity is around 40 to 70 percent. What that means is that of all the variation that we see, for example, in body mass index, 40 to 70 percent of that variation is due to the fact that we differ genetically, and 30 to 60 percent of that variation is due to the fact that we have different lifestyles and live in different environments. It’s pretty high. You could summarize it like 50-50, 50 percent genetics, 50 percent environment. And obesity is very much like at the same level as any other complex disease, such as type 2 diabetes, coronary artery disease, they’re all about like 40 to 70% heritability. What is different from other diseases is that obesity is increasingly an early onset disease, where that is like, it’s unlike type 2 diabetes, we already see obesity at a very young age and often kids with obesity become adults, adolescents with obesity become adults with obesity. Then adults with obesity will have kids with obesity, so it becomes like a self-enforcing cycle. That’s different from diseases such as coronary artery disease and type 2 diabetes. Also, multiple organs are involved in obesity. On either side, on the etiology side, on the causes of obesity, but also on the consequences. People with obesity have multiple comorbidities across multiple tissues and organs. It’s actually a common complex disease that’s influenced by environment and by genetics. My talk is going to focus on the genetics and more specifically is how we can estimate a person’s genetic susceptibility to obesity. Talking about the genetics, we have used genome-wide association studies to identify genetic variants associated with body mass index, with obesity, with body fat percentage, and so on. And the first discoveries were made in 2007. And you may be familiar with the first gene ever discovered. It’s called the FTO gene. And it’s still the poster child of obesity genetics. And if you carry one risk variant of the FTO gene, then your body weight will be about one kilogram more than somebody who does not carry any of these risk variants. If you carry two risk variants, which practically means that you inherit a risk variant from your dad and one from your mom, then you will weigh two kilograms more. In fact, I can tell that 64% of the people here on this call will probably carry one risk variant, and 16% will carry two risk variants. So it’s pretty common. And to discover that FTO locus, we need about 5400 individuals, so not that many. And then over the years, we increased the sample size of these genetic studies through meta analyses. And as we increase sample size, we increase the statistical power and we identify more and more and more genetic variants. We’ve mostly focused on body mass index, but because it’s commonly available, it’s easy assessed in cohorts. That has given us the most statistical power, but variants have also been identified first for other things such as body fat percentage. What we do with these genetic variants is twofold. On the one hand side, we use these genetic variants to gain insight in the biology because variants line near genes and these genes encode proteins or receptors or hormones and so on. That tells us more about the biology that underlies weight gain or body weight regulation. On the other side, we can now start using these genetic variants to, for example, do genetic prediction. Today, I will mostly talk about the genetic prediction, but let me just briefly capture almost like for completeness, what have we learned in terms of biology. Here you see on the slides that we have identified more than 1,700 genetic variants associated with body mass index or some related phenotypes. What we’ve seen is that genes that lie nearby the genetic variants, and I’ll come back to the genetic variant, I’ll explain a little later what it is. But the genes that lie nearby a genetic variant associated with body mass index, their expression is predominantly enriched in the brain compared to genes that do not lie nearby a genetic variant associated with body mass index. This is a nice graph from 2015, now already 10 years ago, but here it shows that the tissue enrichment analysis, the expression analysis clearly shows that the genes near these BMI associated variants seem to act in the central nervous system, probably influencing hunger, satiety and reward. So a major cause of obesity is indeed that some people feel just more hungry all the time, or some people feel less satiated all the time, or some people just feel like eating as a sense of reward. So we know from genetics that obesity is really like a disease in the brain and people are often surprised about it because when I talk about the genetics of obesity, people in the first place will often will think like, oh, it’s like genes that act in adipose tissue. But no, it’s like the driver or the brain that’s driving the hedonic aspects of food intake. But that’s sort of on the broad level, that’s the biology. What’s really that we need to do is translate each of these variants to the actual genes because in this analysis, we don’t look for the specific genes, we just group all the genes within a locus. A lot of activity needs to go into identifying what is the causal gene within each genetic locus. Take here is the FTO locus. Here we have what we call a locus zoom plot. This is a piece out of chromosome 16. FTO is located right here and you see all these genetic variants are highly significantly associated with body mass index, 10 to the minus 60. And for a long time, people believed that FTO is the causal gene. But then after more and more analysis, people started to realize, researchers started to realize that it’s actually not FTO, but genes nearby FTO. And in fact, a few years ago in 2021, there was this paper by Marcelo Nobrega and Melina Klausen, it’s a group, two groups, one from the Broad Institute and one from Chicago, who made sort of a concluding paper where they showed that, okay, here is the variant of FTO located in this area. And within neurons, it seems that this location communicates with genes in this area, IRX3 and IRX5. And also in the pre-adipocytes, it seems that FTO communicates with IRX3 and IRX5. So here the conclusion was that it’s not FTO that is the causal gene, it is genes downstream of FTO that’s the causal gene. But it’s taken them like more than 10 years to figure that out. And that’s just one locus. So a big challenge in our field is like translating these 1,700 loci into like identifying the causal genes. And that’s like work that is ongoing, but it’s not what I’m gonna talk to you today about. Today I want to talk to you about using these genetic variants to assess people’s genetic predisposition. And in fact, I want to talk about a more recent analysis that we’re doing that actually has not been published yet. It is where the Giant Consortium does a meta-analysis. And the Giant Consortium, I may not have mentioned it before, but it’s been driving many of these meta-analyses. And their most recent meta-analysis combines data of 2.1 million individuals, is a collaboration with 23andMe that has shared data of 3.5 million individuals. So now we have 5.6 million individuals into one big meta-analysis. And now we go from 1,700 variants to more than 9,000 variants that are associated with body mass index. And if I just mentioned, like, we need to translate all these identified variants into meaningful genes, it’s like this becomes like a real challenge. And that paper, the actual, what we call the gene discovery paper, that has not been published yet. We’re working on it, we’re finalizing, but it’s been really hard to write a really nice, cohesive and to the point paper. In fact, the Manhattan plot, which is how we summarize these genetic variants, looks like this. So here is our genome going from chromosome one up to chromosome 22. And every dot here represents a genetic variance associated with body mass index. And you see the scale goes here to 300, actually it goes way higher. And the significance level is 10 to the minus eight, so that’s some or 10 to the minus nine that’s somewhere located here on the bottom. So you could say like pretty much every variant is associated with body mass index. So to do the variant to function translation, it is really hard, but using these variants to assess a person’s genetic susceptibility through creating polygenic score, this is amazing. This is very strong and powerful data. So that’s what I’m gonna talk to you about today. And before I explain you what a genetic score is, I wanna tell you what a genetic association is. So first of all, what is a genetic variant? And this takes us back to maybe some classes in biology from undergrad school. So take a person, their genome consists of 3 billion base pairs and there’s only four types of base pairs. There’s A, C, G and T, and that’s about it. Within a gene, these base pairs per three, A triplet of base pairs encodes an amino acid. For example, GGC encodes for glycine, GCC encodes for alanine and so on. And then a series of amino acids, they encode for a protein. Now, one string of our DNA comes from our mom and the other one from our dad. And then if you compare multiple people, take these five people, then you see like a remarkable similarity. It’s like, in fact, only like 0.5% of my genome differs from your genome, but it makes me look and be very different, even though it’s like most of our genome is very similar. But if you look closer here, then you see that indeed there are some variations. Look right here. Some individuals are, it’s like TT homozygous, so I have the T here, and some individuals have an A at the same spot. And you might say, it’s just one genetic variant that doesn’t really matter, but it might matter. For example, in this case, the TCC encodes for serine, but the ACC encodes for threonine. And maybe this faults the protein totally differently. And maybe it’s a receptor that all of a sudden works better or works worse. But this is the type of genetic variants that we study in gene-wide association studies. And of course we don’t study five individuals, we study thousands of individuals. As you saw, we study 5 million individuals. And of course, we don’t study just one genetic variant. We actually study millions of genetic variants these days. And what we do, actually per variant, we group the population into three groups. The TT homozygous in this case, the AT heterozygous and the AA homozygous. So if we take that here, what we then do is we simply align these people’s body mass index and calculate per group what their BMI is. And this, of course, is just an example. We test whether the genetic variants are associated with body mass index. And here you could see that there is a nice relationship with the more A alleles an individual carries, the higher their BMI is. So we simply do a linear regression, assuming there’s an additive effect of each A allele. So this type of associations we do like millions of times. And because we do it so many times, the significance level is so stringent, but this is really like at the core of GWAS. It’s not more complicated than that. The complication comes when we start to interpret it. But now it’s like, we can start building genetic scores. And just a reminder, it’s like, I’ll be talking about the genetic score that we built based on the 5.6 million GEOs. Just to build a genetic score, take as an example, these six individuals and take variant 1. Individuals can score 0, 1, or 2, not more. This means that individuals can carry for this given variant and this variant is a socially polymised index, and that’s why they’re part of a score. They can carry no risk variants, they can carry one risk variant, meaning that they inherited that variant from either mom or dad, and this individual inherited two risk variants from both parents. We look at that for all the variants that are associated body mass index. As I mentioned, we have more than 9,000. In this case, we would combine 9,000 individuals. In fact, even more, but that’s a detail that I’ll spare you from. But the general principle is that we will start counting how many BMI associated risk variants one individual carries. Now, one extra. In the end, we come up with a score by simply adding up how many risk variants an individual carry. An added nuance is that we not just add them all up, we take into account how strong the association of each variant with body mass index is. Like if variant 1 has a very strong association with body mass index, that will make it way more in the score. If a variant 3 has a very small effect, then it will weigh less in the score. Such a score can actually tell you at the individual level what people’s susceptibility is. This contrast to the heritability, when the heritability is at the population level, the polygenic scores is at the individual level. So each individual can be given like a specific score. When we look at the population level, genetic scores are nicely distributed. Most of us will have an average genetic susceptibility. Very few of us will have a very high genetic susceptibility and very few will have a very low genetic susceptibility. So how can we use these genetic scores? We can use these scores at the general population level to assess how much genetics contributes to body mass index at a more, I would say, tangible way. Here we don’t talk about heritability, here we talk about explained variance. For example, here, you see that the score, and I suggest we focus here just on the blue dots. The score that we just developed in our new paper based on the five million individuals explains 17 percent of the variation in body mass index. This is important to remember. With the genetic score, we don’t explain 100 percent. In fact, we also don’t explain everything of the heritability. Remember, the heritability is about 50%. No, we explained 17%. So this is still a caveat, but we believe that 17% is actually pretty good already. And I’ll come back to the comparison with other scores in the next slide. But here, if we now look across other ancestries, the score performs best among European ancestries. Central South Asian performs pretty well. like at mixed American performs really well. So overall it performs pretty well across other ancestries, but never as good as in European ancestries, but it really does not perform that well in populations of African ancestry. And that is actually, that’s what we see is like in obesity genetics, in obesity polygenic scores, but we also see that for many other genetic scores. That may in part be because the samples and the sample size of our studies are predominantly of European ancestry and far less of African, but that’s not the only reason. The African genome has much more granularity, which makes it very different from in particular the European genome to really capture all the variation. If we now compare this score with the most recent of the most commonly used score, we call it the CARA score, then we also see that we perform much better. Here you see in blue, the new score where we explain 17.6 percent, whereas the old score performs below 10 percent. We almost double the explained variance compared to the previous score. We’re really satisfied that even across the board, the new score that we built based on these five million individuals really gave us more power and more accuracy to assess people’s susceptibility. If we now look at the distribution, and this plot is not so easy to look at, but here on the bottom, we have the people’s, it’s not quite the people, we have a scale that goes from 0 percent, the lowest score here up to the right to the highest score. So these are the people who are most susceptible and these are the people at least genetically susceptible. And then here it dropped off, but here it tells you the percentage of people who have obesity, sorry, the percentage of people who have obesity. So if we look here, this is the, here on the right-hand side, these is the people in the top highest percentile of the score. And you see that 70% of people in this top percentile of the score, 70% has obesity. Yeah, and that’s actually remarkable. It’s like, first of all, it’s remarkably high, but at the same time, it’s also remarkable that still 30% who have a high genetic score do not have obesity. Maybe they’re overweight, maybe they’re even normal weight, but it also tells you that the genetic score is not your destiny. On the other side, interestingly, we see that the lowest percentile, actually nobody has obesity. And here in blue, we get the newest score, so we get more extreme on both sides, whereas the older score there, it’s like in the top percentiles, about 55% had obesity. Again, it’s like showing that the newest score is really getting better at identifying individuals with obesity. Now, can we use this genetic score to predict obesity in the future? I’m gonna tell you just a little bit about how we do this prediction. So take a genetic score, we can say that it’s a genetic test. Then we want with that test, the higher the score, so the more risk alleles we have, the more extreme, the higher the people’s risk is to have obesity. On the other side, the fewer risk alleles that you have, we want them to have the lowest risk. Ideally, these two distributions do not overlap, which means that if you have a lot of risk alleles, if your risk score is really high, then you have obesity, and if you’re on the low side, then you do not have obesity. That would be perfect. In that case, the area under the receiver operator curve or the C-statistic would be one. Perfect prediction. Poor prediction is when the score actually does not differentiate these two distributions that they just lie on top of each other. That’s almost as good as flipping a coin. In this case, the area under the receiver-operator curve is 0.5. The prediction will lie between 0.5 and 1. Most of the time, we have a genetic score that will differentiate these distributions but not completely. Let me bring you to the real data. Please do remember that the predictive ability in this case varies between 0.5 and 1, 1 being perfect, 0.5 being flipping a coin. Here, I’m showing the predictive ability here from the new score compared to the old score, the Kera score versus the latest paper. And then you see that to predict common obesity, BMI over 30, we do okay, 0.73 is like not clinically useful, sorry. Not clinically useful, but it’s definitely better than before and definitely better than flipping a coin. The more severe obesity, if we wanna predict the more severe forms of obesity here, like BMI over 35, then it’s not bad, it’s like 0.79. and even more severe, we actually do even better. So the new score seems to be particularly good or getting better at predicting severe forms of obesity. And it’s like, you can imagine that these forms of obesity may actually even be more genetic than the more common forms of obesity where environment might be more important. So this is a snapshot of the data that we presented in the paper. This is the full table. I showed you these results from the UK Biobank predicting obesity class 1, obesity class 2, and obesity class 3. These are the results here in the two columns right here. But you see also the results of Biome and the MVP Million Veterans Program. It’s a little more nuanced. I showed you the best results actually. We perform very well in the European ancestry population of the biome population is in hospital-based population in New York, but maybe not so good in the million veterans population. So it depends on the population where you tested in how well the score performs, but in general, it’s a big improvement compared to last year and also the more extreme the obesity is we predict, the better the prediction will be. Then we already see that our score starts to work early on in life. So here we data from the Allspach studies about 7,000 individuals from the Allspach study, which is a birth cohort in UK. And here we split the population into individuals whose genetic score is in the bottom 10%, in the middle 80% in yellow, and in the top 10% in blue. And then you see already by age five that these three curves start to diverge and that this spread, this diversion actually increases as individuals get older. So now you can start wondering, like, okay, how can I use these scores? Like, if I tell you, like you’re all adults, what your genetic score is, what are you gonna do, right? I tell you, it’s like, you have a high score. Should you be scared? Should you stop eating? Should you start exercising? Actually, you’ve lived with a high score for the past 30 years, 30, 40 years. Like it’s like all of a sudden, should things change or you have a low score, does that mean like, oh, you can start changing life? Actually not. So you could say it’s like, actually at my age, it doesn’t really matter anymore. It’s like, I know my susceptibility and I’ve lived with it and I’ve adjusted my life towards it, right? So when is the score really valuable is really like very early on in life. And that is because different from other diseases, obesity is an early onset disease, right? If I tell you what your score is for lung cancer or for type two diabetes, now at this young age, it may be informative and it may prevent you from doing things, but for obesity, that’s very different. So let me try to explain that in this graph. So like where my argument is that the score is most informative at a young age. So important to remember here is that we’re trying to explain the variation in body mass index at age 18. Yeah, so imagine at age 18, actually a very good predictor of your body weight at age 18 is actually your body weight at age eight. Your actual body weight at age eight or your body mass index at age eight, 10 years earlier. Your body weight at age five is also still a good predictor for your body weight at age 18, but it’s not as good as at age eight. And the further you remove, you are removed from age 18, the less well the actual body mass index will be predictor of your body weight at age 18. And this is actual observed body mass index, right? So if we go back and now add genetics. So I know my body weight at age eight, so I sort of can predict what my risk of obesity is at age 18. Does knowing my genetics, my genetic susceptibility add anything? Just a little bit, right? Not meaningful, you could say. But actually, if you go back in time, then here where your actual body weight at age five does not explain as much, there the genetics seem to have a more meaningful contribution. It’s particularly at these earlier ages that knowing your genetics, you have more information to assess your risks later on in life. This is also one of the key outcomes in the paper. This genetic score has the most information in terms of prediction of obesity early on in life. In fact, at an age where body mass index is not as predictive towards your later health. Then also emphasizing that body weight is interact or your genetic score interacts with your environment. We show the results of trials, weight loss trials, the diabetes prevention program and the look ahead studies to weight loss interventions. Here we see that the higher your genetic score is, and this is the meta-analysis, the greater your weight loss. That was a surprise to us. We thought that people with a high genetic score for obesity would lose less weight, but here we show that they actually lose more weight than than people with a low genetic score. However, after the intervention, when people are back actually gaining weight, then we also see that people with the highest genetic score actually gain the most weight. So it seems that our score captures very well the gene environment interactions. It’s like the highest score shows you that you lose a lot of weight, so very responsive to the environment, then you gain a lot of weight as well. In our context, we’re living in a basic genetic environment and it’s particularly people with high genetic score who will respond to this changing environment and in our context will gain weight. Just to add that. So far, I’ve told you everything about genetic score for body mass index. The next set of slides, I wanna talk more about genetic score for subtypes of obesity, because we all recognize that body mass index is not that good of a phenotype. It only captures like one part of body size and it also doesn’t capture, it doesn’t capture the heterogeneity that goes with obesity. So in a more recent paper, we decided to identify subsets of obesity using genetics. And here I’m just first showing you the general principles. Like take all these genetic variants that are associated with body mass index. That’s a starting point, but not all these variants are in the same way associated with other variants. So these variants are all associated with body mass index by definition, but some of these may indeed also be associated with waist-hip ratio, some are not, and some are maybe associated with lower waste depreciation. And at the same level, some may be associated with fasting glucose levels, higher fasting glucose levels, but other are made lower fasting glucose levels and so on. So we can actually, if you add more descriptive phenotypes to your genetic association analysis, then you actually start to see subsets of genetic variants. For example, and this is all just like schematic, just to explain the concept. So this is a set of variants in cluster one. You could say that are all associated with higher BMI, higher waist depression, higher glucose. But in cluster two, now you get variants associated with higher BMI, higher waist depression, but lower glucose levels and so on. So here now you start to identify genetic clusters among like all variants associated with body mass index. And in a way you could say is like, if we can identify these clusters, so now we can calculate these six clusters into a population and identify people who score high for cluster one, for cluster two, for cluster three, and so on. And maybe these subsets of clusters would benefit for different kinds of prevention and intervention strategies. So this is just the concept, the idea. And this is something we, my team are working towards. And one paper towards this general idea of genetic subtyping was recently published by Natalie and JJ. And what they did is they really wanted to focus on genetic uncoupling. So internally, we call it the Healthy Obesity Project. Externally, we call it the project that was obesity is uncoupled from its comorbidities. And we use data of the UK Biobank, 452 individuals. We have three metrics of body size, BMI, body fat percentage and waist-dip ratio. And we have like eight cardiometabolic risk factors, cholesterol, triglycerides, blood pressure, glucose and insulin. And then per analysis, we make a new phenotype. We make a phenotype where we, So the principle is like take body mass index. It’s a right skew distribution, somewhat right skew. LDL cholesterol is also somewhat right skew, but now we transform them to be both each of them on a normal distribution. We do an inverse normal transformation and each of them have now the mean of zero and a standard deviation of one for both. And now we subtract one from the other. So now we get a new score. So if my BMI is high, I will have a standard deviation score for BMI of plus three, but if my LDL cholesterol is low, it will be minus three. So now if I subtract one distribution of the other, my score will be, depending on how you do the subtraction, will be plus six. Plus six meaning that I have healthy obesity. It’s like I have obesity, but with a very low LDL score. And this way you get, and on the other side, you have minus six where you have individuals who do not have obesity, but have a high LDL score. So you get a new distribution for this pair of phenotypes, and then we can do genetic discovery studies for this new phenotype. And we did that, we did that for BMI and LDL cholesterol, but we did it for all these possible combinations. So we did 24 new genome-wide analyses for these new healthy obesity phenotypes. And then we get a lot of new data. And we look like how the variants overlap with each other. And we identified 205 loci within 266, with 266 independent association signals. The bottom line is we can now use these 266 association signals to make a new genetic score. And this time, this score will be a healthy obesity score. it will be associated with higher adiposity, but with lower risk of comorbidities. That’s what you see here. We validated it here. Here’s the radar plot and you see the null line is right here in the middle. In red, you see the uncoupling score and you indeed see that it is associated with higher hip, waist, body fat percentage, fat-free mass index, BMI, gynoid fat, android fat, and so on. But it’s also associated with lower visceral and subcutaneous fat distribution and with lower liver fat fraction compared to the obesity score. and the obesity score or the unhealthy obesity score is here in blue, that clearly has a stronger effect, particularly when it comes to fat distribution. I’m sorry, I’m aware I did not explain that very well, but what we contrast here is the uncoupling score that represents the unhealthy, though that represents the healthy obesity, clearly showing a healthier fat distribution compared to the unhealthy obesity score in blue. And also when we look at the cardiometabolic phenotypes, now the null line is right here. The unhealthy obesity score lies on the outside with higher blood pressure, higher glucose, higher HPOMC, whereas the healthy score lies on the inside with lower levels, except of course for HDL, which we expect to be higher when it comes to health. So now we have like a nice score that captures obesity, but the healthy obesity. And if we do an analysis where we take into account the medical records of all participants, then we see nicely that both scores, the healthy and unhealthy obesity score are both associated with obesity, maybe the unhealthy even a little more, but that the healthy obesity score has protective effects on what we call disorders of lipid metabolism, type 2 diabetes, essential primary hypertension, and so on, compared to the unhealthy obesity score that is associated with higher detrimental effects. Importantly though, it’s like even though we call it the healthy obesity score, we should call it the metabolically healthy obesity score because when we look at diseases that are more the result of weight bearing, then we see that even the unhealthy obesity score actually is also associated with higher risk of say sleep disorders. emphysema, varicose veins, and so on. So even though we captured a metabolically healthy obesity very well, it does not mean that obesity is really healthy in the context of weight bearing diseases. Here we also replicated that in children, which is always nice that you see the genetic effects are already there early on in life. This is data of 11, 12-year-old children of the Holbeck study, 3,500 individuals. The right-hand side is actually even more telling, oops, where you see that again, the blue score lies more at the outside of the null line here. Oops, sorry. And the red score lies on the inside. So it’s like the red score already showing early in life that the higher you score on this healthy obesity score, the more protected you are, even though there is still an association with higher adiposity. So this is just to say that the genetic score for obesity is great, and it can predict obesity very well, but it does not give you much resolutions, like it predicts any obesity. And here in this new study, we tried to disentangle in this example, the healthy from the unhealthy obesity. And in my group we were working to get even more granularity in scoring individuals’ genetic risk. So how can we now implement these genetic scores in clinical practice? So I see these genetic scores mostly as biomarkers. And remember, these scores explain not even 20% of variation in body mass index. So they are definitely not your destiny, right? They give you a level of risk. They give you an idea of what your risks are. And here in Denmark, we have this app that connects you with the healthcare system. And it tells you about meetings with the general practitioner, the medication that you’ve been prescribed, the tests that you have done. So maybe you have an HbA1c test or a cholesterol test. So the way I imagine it is, is that we get all these tests in our records And maybe at some day we will have the results of our genetic testing. And genetic testing, it sounds very scary because we always connect it to like monogenic forms of obesity and like congenital diseases and or monogenic forms of disease and congenital diseases. But in this case, it would just be a biomarker. So on the one hand side, we get a medical record that tells you about your weight and your height and your blood pressure and your lipid levels. And on the other side, you get information about your genetic risk. And it should be interpreted almost in the same way as your lipid levels. So here it tells you that you’re genetically predisposed somewhat more than other people to obesity. And it doesn’t mean that all of a sudden it’s like you have to panic. Because as I said, it’s like genetics is not your destiny. It’s like part of your risk for disease, but it does not mean that you therefore will get the disease. And as I said, it’s like, once you have your genome tested, you can make a genetics score for obesity, but genetics scores also exist for many other diseases. So it gives you a feel of, it’s like what your genetic susceptibilities are to certain diseases. And like, I can imagine like it could be presented like this. This of course requires a lot of education, not only or not the least from healthcare practitioners, but also from the populations and how to interpret these diseases and how to deal with it. Just to finish my talk here, I wanna bring back an old study actually, but I find it’s a very telling study. This is about what’s the impact of informing people about their genetic risk. And this study was published, caught 2011 already, it’s like 14 years ago, almost 15 years ago in New England. And they did a study where they informed individuals about their genetic risk, and then they followed them up for three years. And then in a further follow-up study, they followed them up for another year. And they did genetic profiling in that time, and they informed individuals about their genetic risk for 23 diseases. There was 3,600 individuals involved. There was quite a bit of dropout. They followed them up for three months and then again for one year. And they tested about their anxiety. So informing them about the genetic risk, did it increase their anxiety? Did it change their dietary fat intake? And did it change their exercise behavior? So they look at zero, three years, three months and one year. And interestingly, it’s like overall, the short term, three months, like overall, it did not in fact increase people’s anxiety or didn’t change their lifestyle. And that’s overall for the 23 diseases. But if you look into the supplementary data, then you could see that in fact for obesity, there was the only significant effect in those tables. And what we saw here interestingly is that individuals who were told that they were genetically predisposed obesity, they increased their fat intake. Mostly think probably like my argument is like me, they probably says like, that’s it. I have it. I have the genes like, I cannot do anything about it. So why is like, waste my time? And it’s like, just let’s enjoy life. That was an immediate response. After one year, actually, that effect on dietary fat intake had disappeared. So that increase in fat intake seems very acute. So just to say that informing genetics, informing people about their genetic risk may differ from disease to disease. If I tell you now about your risk of type 2 diabetes, which is in the future, maybe you will change your behavior to make it healthier. But if I tell you about your genetic susceptibility to obesity, if you’re not very well informed, you might say, “Well, that’s it because we are in it basically. it’s not a disease that’s going to happen in the future. Anyway, just a few, just a summary, we know that hundreds, actually thousands of genetic variants are associated with body mass index. The more variants, the stronger the samples, the bigger the sample size, the better the genetic scores are that we can make. These polygenic scores already act early on in life, and in fact, we could say that they are most informative early on in life. Important here is a message is also that these should be considered as biomarkers, as a genetic risk of obesity, not as your destiny. Genetic scores can help identify subtypes of obesity. Like I showed you one example, but there’s definitely more work that can be done in that direction. Of course, translating these genetic risk into the general population, into clinical care, will take still a lot of education from all parties involved. And I’ll finish here. I’m very grateful for my team here at the University of Copenhagen. Also still a small team at the university at Mount Sinai Hospital. And of course also for the funding. And I just don’t want to forget the many authors that were involved in the study that we published in Nature Medicine. And I’m happy to take some questions if we have time.

Speaker 1
53:05 – 53:36
– Yes, thank you so much, Professor Loos. Amazing presentation, I think really highlights how exciting obesity research is at the moment. So we do already have several questions. and if you do have questions, please do use the Q&A function, but we’ll kick off something to ask you. So one of the questions is, can you see the polygenic risk score being used to support public health intervention aimed at reducing the risk of obesity development? And I’d like to add onto that. Do you foresee any logistical or potentially even ethical challenges?

Speaker 3
53:38 – 55:11
– Yeah, there’s a lot of ethical challenges could be at different levels. Typically as researchers, we are more conservative than the actual public in the sense that we think that the general patient or public may not want this kind of genetic information, whereas it’s like we know from surveys that they actually do want that kind of information. But it’s indeed true, like we think particularly in the US where it’s like insurance companies are like strongly involved in the healthcare. if they come to know about your genetic predisposition, which is basically you can determine at birth, maybe they can decide in these like, okay, it’s like you have that genetic predisposition where you’re just like, kind of stop, you cannot be our patient or you cannot be our customer. So indeed, there are some of these kind of ethical challenges in that sense that we may not want to share this genetic information with the broader like healthcare system. But I would think that if the right checks are in place, like I think that could be dealt with. It’s almost like, do we not exactly, but to some extent, it’s similar to like, do we want insurance companies to know our blood pressure and so on. So it’s as I said, like, I really wanted to be interpreted as a biomarker of, of your risk of

Speaker 1
55:11 – 55:29
disease. Thank you. And sort of staying on the theme of intervention. There was a question around differential response to obesity management medications and whether you have any insights on that.

Speaker 3
55:29 – 56:44
Yeah, that’s a developing field. I was part of a paper that showed that the higher score and it was using the old score does not affect how people respond to the weight loss medications, the current weight loss medications. People with monogenic variations in MC4R, is like a commonly, is like a gene known to cause monogenic forms of obesity, does not seem to affect your weight loss. Then there are other studies that do show that a coding variant within the GLP-1 receptor itself seems to affect how people respond. I would say that it’s a developing field and I would love to analyze that kind of data, but currently I have not been able to access. It seems like particularly the larger studies, because that’s what we really need, like a nicely controlled trials could help indeed provide these kind of insights. And even we could work the other way around, like identify the non-responders and try to identify, it’s like, why are they not responding? And identify the genetic of these people? Yeah, really interesting field.

Speaker 1
56:45 – 56:58
Really, really exciting. And on that point, something that I wanted to ask was around the differential effect between ancestries and whether that extrapolates on on the response to an intervention to?

Speaker 3
56:59 – 57:18
I can imagine. I don’t know whether two are related, but response to weight loss medication is different across ancestries but I don’t think that’s necessarily related to the genetics. Yeah. And another question in the last three minutes

Speaker 1
57:20 – 57:40
was around, I think what this question is asking is around the individuals with a very high polygenic score who don’t have obesity and insights maybe a little bit more deeply into what’s happening in those individuals? Is it environment phenotype driven or do you have any

Speaker 3
57:40 – 58:08
any more to share? Not yet. So I am doing in at Mount Sinai, I am doing a study where I identify individuals with a high genetic score who do not have obesity and then we bring them in the metabolic rooms, like we very much control their diets, their exercise, they do some exercise, so we do the very deep phenotyping to really see like why are they resilient to their genetic predisposition. Yeah, interesting field as well as like interesting individuals. Yeah.

Speaker 1
58:08 – 58:23
Really interesting. And probably the final question is, would you be able to explain a little bit more around the term granularity with regards to African genotypes?

Speaker 3
58:25 – 59:43
So the European genome, basically, it’s like in general terms, consists of bigger blocks, meaning that these are genetic variants that basically are inherited as a bigger block, whereas the African genomes, like which you could say is older, there we have like these blocks are smaller. So that means that, yeah, I feel like if we make a score that is predominantly actually based on populations of European ancestry or some South Asian, Hispanic as well, it’s based on bigger blocks. So that if we do build the same kind of score with African genome data, we don’t actually with African genome data, we don’t capture all the information that is actually there. not so good of an explanation, but in part related to the differences in the genomes, meaning mostly in terms of how the blocks are inherited, and in part because the sample sizes we have for people of African ancestries are still lagging behind that we have for other ancestries.

Speaker 1
59:43 – 01:00:08
Right, really interesting. Thank you so much. Unfortunately, that’s all we have time for in There are plenty more that we could have asked you. But again, I just wanted to extend a huge thank you for your time and expertise to talk to us about this really, really exciting area for research. So thank you very much, Professor Loosom. Hopefully we’ll see the rest of the Early Career Network at our next webinar.

Speaker 2
01:00:09 – 01:00:12
Thank you. Thank you so much. Bye. Thank you so much.