This episode examines a large-scale study that maps how genetic sex differences are distributed across the genome for 157 quantitative traits in the UK Biobank. Using LAVA and complementary methods the authors test local heritabilities, local genetic correlations, and equality of genetic effects on raw and standardized scales to reveal locus-specific sex-dimorphic signals that are often masked by genome-wide averages.
0:03Yeah. Yeah. Yeah. Yeah, uh, yeah. Welcome to Base by Base, the paper cast that brings genomics to you, wherever you are. Thanks for listening, and don't forget to follow and rate us in your podcast app.
0:29We know that males and females have different disease risks and average biological traits. Right. I mean, that is just a basic fact of human biology. Exactly. But what really happens when we stop looking at the genome as one giant blurred average, and, you know, actually zoom in on specific neighborhoods of our DNA.
0:47It's huge perspective shift. It really is. Think about it like a giant mosaic. If you stand far back. A mosaic might look exactly the same for men and women, the overall picture, the broad strokes, they seem completely identical.
1:00Yeah, you just see the overarching pattern. Right. But if you walk right up to it and examine the individual tiles, are the colors actually arranged completely differently? How could this change the way you understand your own genetic risk?
1:12That shift in perspective, you know, from the macro to the micro is really the foundation of our deep dive today. For a long time, the scientific community just evaluated that mosaic from across the room.
1:23And today, we are finally analyzing the individual tiles. Today, we celebrate the work of the team at Bri Universitate Amsterdam and the Amsterdam University Medical Center, who have advanced our understanding of the genetic architecture underlying sex differences.
1:37They really have. It's pretty groundbreaking work. Totally. So to give you a clear overview of our mission today, we are untacking their recent nature communications paper. They leverage data from over 360,000 individuals in the UK Biobank, which is massive.
1:53It's a huge data set. Right. And they analyzed 157 distinct quantitative traits. We're talking everything from blood biomarkers to physical measurements. Our goal in this deep dive is to understand how analyzing the genome at a highly localized level reveals profound biological differences between the sexes.
2:12Differences that, honestly, global genetic averages have completely hidden from us until now. Yeah, hidden in plain sight. Exactly. And to appreciate the scale of this paradigm shift, we really need to look critically at the historical approach.
2:23Historically, genome white association studies or GWS, they basically just group males and females together. Just lumped everyone into one massive cohort. Yeah, exactly one large cohort. They would look at a trait, like, say LDL cholesterol, and calculate the genetic associations across the entire population.
2:42And they'd include sex merely as a statistical covariant. So they just adjusted for it mathematically. Right. But that mathematical adjustment treated sex as a simple baseline shift. It fundamentally masked any divergent, you know, sex specific genetic mechanisms that were actually at play.
3:00Okay, let's unpack this. Wait, so if a gene passes the threshold for statistically significant in men, but just barely misses that threshold in women, researchers used to assume that meant a biological difference.
3:12But the difference between significant and not significant isn't always significant itself, right? This raises an important question. Because that exact statistical trap, combined with global averaging has hidden true localized biological differences for years.
3:26Wow, really? Yeah. When the field realized that throwing everyone together caused antagonistic effects to just vanish, they shifted tactics. Researchers began running sex stratified studies. Meaning they'd run one GWS for men and a separate one for women.
3:39Exactly. But that introduced a dangerous statistical trap you just mentioned based on arbitrary significance thresholds. The underlying molecular mechanism was often functioning identically in both sexes.
3:51But because of slightly different sample sizes or like data noise. One hit the threshold and the other didn't. Right, which created this totally false narrative of sexual dimorphism. So to correct this, the field evolved again.
4:04They move toward calculating what we call global genetic correlations. Meaning they started comparing the overall genetic architecture of a trait between sexes across the entire genome, but calculating a single correlation coefficient for the entire genome just creates a new averaging problem, doesn't it?
4:21It absolutely does. If one genetic pathway has a strong positive correlation between sexes, and another pathway has a strong negative correlation. The global average just flattens it out. You completely lose all the topological detail of the genetic landscape.
4:34Precisely, which brings us to the core methodology of the Vraillet University study. They bypass global averaging entirely. Okay, so how do they actually do that? Well, instead of looking at the genome as a single entity, they partition the autosomal genome into 2,495 semi-independent blocks, and each block is approximately one megabase in size.
4:55Let's clarify why one megabase is the magic number here. It comes down to linkage disequilibrium, right? Our DNA isn't inherited as completely independent letters, it's inherited in chunks. Right, exactly.
5:08those chunks are key. So these one mega-based blocks reflect those natural boundaries of recombination. It allows us to analyze localized genetic covariants without signals bleeding too heavily into adjacent neighborhoods.
5:19That is the exact biological rationale. Yeah. By isolating these winkage to equilibrium blocks, they used a statistical framework called LAVA. It stands for local analysis of variant association. LA VA.
5:33That's a great acronym. It is, right. So LVA allows researchers to estimate the local genetic irritability and the genetic covariance between sexes within just that specific one megabase window. But the critical innovation here isn't just zooming in spatially, is it?
5:46No, it's not. The real innovation is that they evaluated the equality of genetic effects on two distinct mathematical scales. They use the Raw scale and the standardized scale. So what does this all mean?
6:00Is the raw scale like measuring a car's speed an exact miles per hour while the standardized scale is like measuring how fast it's going relative to the speed limit of that specific road? That is a brilliant analogy, yes.
6:12Only the standardized scale tells you about true heritability. Why is that? Because males and females often have completely different natural ranges and distributions for biological traits. If men and women have different raw genetic effects, it might just be because one sex has a naturally wider variance in that physical trade overall.
6:31Right. If females have a much wider variance in a specific biomarker, an absolute raw increase of, say, 5 units is just a tiny ripple for them. Exactly. But for males, operating with a much tighter natural variance, that exact same 5 unit increase is a massive spike.
6:47It changes the whole picture. So the standardized scale scales the genetic effect relative to the total variance. It gives us the actual proportion of the trait's variation driven by genetic. Precisely.
6:58And applying this dual scale localized approach to those 157 quantitative traits yielded some incredible results. Out of the 157 traits analyzed, 151 exhibited at least one specific genetic locust that differed significantly between males and females.
7:13Wow. 151 out of 157? Yeah, almost every single one. That forces a complete rethink of how we interpret Cleoctopy. We aren't just talking about a handful of sex specific hormone pathways anymore. No, not at all.
7:26We were talking about nearly every continuous trait in the human body operating differently, depending on the biological sex at a localized level. Blood biomarkers, specifically urate, lipids, and testosterone show the highest concentration of these differences.
7:38Here's where it gets really interesting. Let's talk about the body mass index, or BMI, paradox. Oh, this is the most striking example of why global averages fail. Globally, across the entire genome, the genetic correlation for BMI between males and females is 0.93.
7:53And a correlation of one.0 means perfect alignment. So .93 tells us that from a macro perspective, the genetic blueprint for human body mass is essentially identical regardless of sex. Right, and if you rely on global GWAs, the inquiry just stops there.
8:09But using LOV, the researchers isolated Ligus 1727, located on chromosome 11. Okay, what happens there? In that specific one mega-based neighborhood, the genetic correlation between sexes actually plummets to -12.
8:23So you're telling me that overall, the genetic blueprint for BMI looks exactly the same, but if we zoom in to this one specific neighborhood on chromosome 11, the genetic effects are actually pulling in opposite directions for men and women.
8:35Exactly. A specific genetic variant in that block that predisposes a male to a higher BMI, is simultaneously associated with a lower BMI in a female. That is wild. The global average of .93 completely obscured this intense localized biological conflict.
8:50It really did. BMI shows how a high overall correlation hides localized conflict, but testosterone demonstrates the exact inverse of that. Right, because globally, the genetic correlation for testosterone between males and females is strictly zero.
9:04Exactly zero. Statistically, that suggests that genetic architecture driving testosterone in males has absolutely no overlap with the architecture and females. They look like entirely independent biological systems.
9:17But locally, it's a completely different story, isn't it? Oh, completely. When they applied LAVA to the one mega-based blocks, the landscape was incredibly volatile. Some blocks are strongly positive like.
9:28Locust 963 is .85. Indicating a highly shared genetic architecture in that specific spot. Right. But then others are strongly negative. Like Locust 2020 is negative .67. You have these independent genetic neighborhoods actively fighting each other.
9:43And when you average that localized chaos across the entire genome, it just mathematically cancels out to zero. Exactly. It's mathematical illusion, basically. But knowing a genetic neighborhood is acting differently is one thing.
9:55A one megabase block still contains multiple genes. How did the researchers pinpoint the specific biological levers being pulled? Ah, for that, they used a gene prioritization framework called flames. Flames integrates functional genomic data, like how DNA physically loops and folds to bring regulatory elements into contact with target genes.
10:15Which is vital because a regulatory enhancer could be 1000s of base pairs away from the actual gene it controls. Exactly. So flames helps map the statistical signal to the actual functional gene. And they applied this to an overlapping locus for testosterone, specifically targeting the AKR1c gene family. And what did they find?
10:34They found that at the exact same genetic locust for testosterone, the actual causal gene is likely AKR1c2 for females, but AKR1c4 for males. Oh, wow. So the statistical signal is flagging an entirely different enzymatic pathway, depending on the sex of the individual.
10:50Right. The biological mechanism is completely divergent, even though it's the same neighborhood. What's fascinating here is how SHBG, which is sex hormone binding globulin and HDL cholesterol, perfectly illustrate why we need those 2 scales we talked about earlier.
11:04The raw and standardized scales. Exactly. HDL showed significant differences between men and women on the raw scale. The literal change in lipid concentration was measurably different, but when they evaluated the standardized scale, that difference entirely vanished.
11:20Because females just naturally have a larger variance in HDL, right? Right. It's roughly one. 5 times the variant seen in males. So once you scale it against that broader natural distribution, the relative genetic impact, the true heritability is identical.
11:34Which means the raw scale alone would have generated a massive false positive. Exactly. But SHBG is the exact inverse. And arguably one of the most critical findings in the paper. They look at a gene called JMJD1C, on the raw scale, the genetic effects were practically indistinguishable.
11:52But let me guess. The variants for SHPG and females is massive compared to males. Massive. roughly 3.6 times larger. So that identical rogenetic effect represents a minuscule fraction of the total phenotypic variants in female.
12:05But for males, operating with a much tighter natural variance, that exact same absolute raw effect dries a massive percentage of the traits variation. Right. As a result, that gene, JMJD1C showed massive standardized differences, making it three times more heritable in males.
12:21Three times. That completely redefines how we calculate genetic risk. The exact same allele holds triple the predictive weight, just because of the biological environment it resides in. It's staggering when you really think about it.
12:34And it brings us directly to the clinical implications, especially regarding disease risk. Let's look at the APOEG. Which is famous for its role in Alzheimer's and cardiovascular disease. Exactly. The researchers found APOE has twice the heritability in females for LDL cholesterol.
12:50The directional pattern is the same, but the magnitude is vastly different. Double the heritability. That strongly suggests APOE is interacting differently with male versus female environments or hormones.
13:01Right. Like maybe interacting synergistically with circulating female sex hormones like estrogen. This is critical because elevated LDL drives cardiovascular disease. And the APOEE for allele is the strongest known genetic risk factor for late onset Alzheimer's.
13:15Which could eventually explain sex-based differences in cardiovascular and Alzheimer's risk, since women make up nearly two-thirds of the Alzheimer's patient population. Exactly. We are finally moving beyond just observing that risks are different.
13:28We're mapping the specific genetic coordinates where those differences originate. It's an incredible analytical leap, but, you know, we always have to talk about study limitations. These blocks are one mega base wide, which means they sometimes contain multiple genes.
13:43Right, making exact pinpointing tricky without extra tools like flames. Plus, the data is limited strictly to quantitative traits. They didn't look at binary diseases like spizophrenia or type 2 diabetes.
13:55And the UK biobank data set is restricted to individuals of British ancestry. We know that linkage patterns and environmental interactions shift dramatically across diverse global populations. Yeah, replicating this localized analysis across diverse global biobanks is absolutely imperative.
14:11We can assume these specific amorphisms will map identically everywhere. So bringing it all back to the listener. If I'm a patient getting my genome sequence tomorrow, does this mean my doctor needs a completely different instruction manual to read my risks compared to someone of the opposite sex?
14:26If we connect this to the bigger picture. No, we don't need entirely separate manuals. In fact, running totally separate GW as for men and women, halves the sample size and ruins statistical power. Because most LISIs actually don't differ, right?
14:41Right. The vast majority of the genome doesn't exhibit dimorphism. Instead, future prediction models just need to directly account for these specific localized sex interactions where they actually exist.
14:52Like using conditional logic. Calculate baseline risk. But if the patient is female, apply a multiplier to the APOE locus for LDL prediction. Exactly. We maintain the massive statistical power of a combined cohort, but we program the models to be smarter locally.
15:07So, global genetic averages basically mask localized biological realities. By zooming into specific genetic neighborhoods, we discover that the genetic architecture for almost every quantitative trait operates differently between males and females in at least one area, completely changing how we view heritability.
15:26It's a fundamental shift in genomics. It really is. What does this mean for the future of precision medicine when the exact same genetic variant might pull a biological lever twice as hard depending simply on your biological sex?
15:41That is the $1000000 question for the next decade of research. This episode was based on an open access article under the CCBY 4.0 license. You can find a direct link to the paper and the license in our episode description.
15:54If you enjoyed this, follow or subscribe in your podcast app and leave a five-star rating. If you'd like to support our work, use the donation link in the description. Now, stay with us for an original track created especially for this episode and inspired by the article you've just heard about.
16:07Thanks for listening, and join us next time as we explore more science based by base. Yeah. Yeah. Yeah, uh, yeah. Same sky, letters, same long cold view, but the signal bins dependent who's coming through. In the 1000000 base window, the truth gets loud, some light shit different in the day, the crowd.
16:42Zoom in close where the averages break. Rob is standardized, watch the mean and shift shape. If variance moves, the headline can lie. So we test the scale before we testify. Not one genome, 2 stories in the light.
16:55Locust by Locust, the numbers don't align. Sometimes they mirror. Sometimes they collide. Sex and the architecture hidden inside. Read it in the low. Don't blur it in the hole. Different effect sizes Different kinds of control.
17:13Yeah. Across a 100 traits, it keeps showing up. At least one region that won't match the cut. Local heritability, stronger here than there, and power toast the finest when the samples aren't fair. Some correlations flip below the line, like opposite echoes in the same bloodline.
17:29You can't find a story with the blurry frame. One locust can hold more than one name, still clues, stay sharp when you follow the flame, hormone, wire, switches, lip, shift, and aim. So build your models with a sex, so where have you?
17:43Or Mr. Biology? Staring back at you. Not one geno, two stories in the light. Locust, by locust, the numbers don't align. Sometimes they mirror, sometimes they collide. Sex and the architecture hidden inside.
18:01Keep your eyes on the scale. Let the test be the guy. Ooh, ooh. From all the standardize. Let the truth decide.
