Episode Transcript
[00:00:00] Speaker A: Foreign.
[00:00:14] Speaker B: Welcome to Base by Base, the papercast that brings genomics to you wherever you are.
We're diving into, well, a genuinely stressful clinical scenario today. Imagine the weight, just, just for a second, of getting a genetic test result back and it flags a variant in the TP53 gene.
[00:00:31] Speaker A: Yeah. For patients, for their families, that finding it immediately points to liferoom and E syndrome LFS. Yeah. And historically that's meant, you know, a near 100% lifetime risk of cancer.
[00:00:43] Speaker B: It really is the definition of high risk, isn't it?
[00:00:45] Speaker A: Absolutely. For what we call a standard pathogenic variant in TP53, the clinical path is, well, it's intense and immediate. Patients face lifelong annual surveillance. Think whole body MRIs starting incredibly young.
[00:00:59] Speaker B: Necessary, absolutely necessary. But it completely changes a person's life, radically impacts it.
[00:01:03] Speaker A: Yeah.
[00:01:03] Speaker B: So here's the tough question, the one that's bothered clinicians for years. What if the specific variant found in this, let's be honest, catastrophic gene, doesn't actually carry that maximum risk?
[00:01:12] Speaker A: Right. What if it means a lower risk?
[00:01:13] Speaker B: Yeah.
[00:01:14] Speaker A: Or maybe a moderate one. These are the reduced penetrance variants, RP variants.
[00:01:18] Speaker B: And that possibility, it creates this terrifying tightrope walk for doctors making the classification.
[00:01:25] Speaker A: It really does. Because if they get it wrong, say they call that RP variant benign, or maybe a VUS variant of uncertain significance.
[00:01:34] Speaker B: Which happens a lot.
[00:01:35] Speaker A: It does. Then the patient misses out on screening, tailored screening they might really need, and they could end up developing an early onset cancer that maybe, just maybe could have been caught.
[00:01:44] Speaker B: The. The other side of that coin is just as bad, really.
[00:01:46] Speaker A: Exactly. If they over classify it, call it standard, high risk, pathogenic, then that patient might face decades of incredibly intense, expensive, emotionally draining screening that might actually be.
[00:01:58] Speaker B: Overkill for their real risks.
[00:02:00] Speaker A: Precisely. So the community, you know, urgently needed a clearer way, some quantitative definition for this, this middle category.
[00:02:07] Speaker B: Yeah, we needed like a dimmer switch for TP53 risk, not just a stark on off button. Okay, let's unpack this.
[00:02:13] Speaker A: So today we're really digging into the crucial work. Christina Fortuno, Marcy E. Richardson and their colleagues, they came from several places including qimr, Berglofer and Ambry Genetics. And what did they achieve?
[00:02:26] Speaker B: They tackled this really nuanced challenge head on. And they seriously advanced our understanding of how to classify these TP53 variants by actually finding the quantitative features, the numbers that separate these reduced penetrance variants out.
[00:02:43] Speaker A: It sounds straightforward, but the challenge they were up against is almost structural, isn't it?
[00:02:47] Speaker B: It is the main guidelines we all Use in clinical genomics. The ACMGMP criteria, they're fundamentally built for a yes, no answer. Benign or pathogenic.
[00:02:56] Speaker A: Great for the clearer cases, brilliant for those.
[00:02:59] Speaker B: But they really struggle when the biology isn't black and white, when it exists more on a spectrum.
[00:03:04] Speaker A: And even the more specialized guidelines have trouble here.
[00:03:06] Speaker B: That's right. Even the really authoritative gene specific guidelines from groups like Klingon's variant curation expert panels, the VCEPs, you know, they exist for high risk gen like TP53, BRCA1, but even they are kind of tuned to expect that usual high penetrance. So if a variant pops up that looks atypical or seems to have reduced penetrance compared to that high standard, the guidance can get a bit fuzzy. And fuzzy often means VUS variant of uncertain significance.
[00:03:34] Speaker A: Exactly. And misclassifying one of these reduced penetrance variants as a vus, well that creates a massive clinical blind spot.
Labs have limited resources. VUS cases often get pushed down the priority list for re review.
[00:03:47] Speaker B: Leaving patients and their doctors just stuck in limbo.
[00:03:50] Speaker A: Yeah, precisely. So maybe let's just quickly recap what standard liferometi means. It's caused by these germline pathogenic TP53 variants, right? And it's known for these early onset core cancers. Think breast cancer, brain tumors, sarcomas, adrenocortical carcinoma. Often showing up really early, before age.
[00:04:10] Speaker B: 30, sometimes even in childhood.
[00:04:11] Speaker A: Exactly. Standard pathogenic variants carry those very high risks very early on.
[00:04:16] Speaker B: Okay, so the study's mission then was basically to define that middle path.
[00:04:20] Speaker A: That's it. They aim to look at a whole range of characteristics. Functional data, bioinformatics, something called immunogenicity, frequency data, and of course, clinical outcomes.
[00:04:28] Speaker B: All to build a solid profile that could reliably separate these RP variants from the standard pathogenic, the P variants and the truly benign the B variants.
[00:04:38] Speaker A: They started by building their datasets. First they combed through Clinvar submissions. They were looking for any variant descriptions that used terms like reduced, moderate or lower penetrance, lower risk. That gave them their initial pool of potential RP variants.
[00:04:52] Speaker B: Okay, makes sense.
[00:04:53] Speaker A: Then they needed solid ground truth. So they set up two really robust reference sets. They got 62 missense variants everyone agreed were benign, and 113 missense variants everyone agreed were pathogenic based on VCEP decisions or, you know, multiple labs saying the same thing without conflicts.
[00:05:10] Speaker B: So those were the groups to learn from the B and the P. Exactly.
[00:05:13] Speaker A: The training sets. And the next step was just intensive analysis. They took Every variant B, P and potential rp and ran it through nine different metrics across four main areas.
[00:05:23] Speaker B: Okay, let's break that down. Sounds complex.
[00:05:25] Speaker A: It was comprehensive. First function, how well does the p53 protein actually work when it has this specific mutation? You know, P53 is called the guardian of the genome for a reason. They use four well established functional assays. Tools the TP53VCEP already uses, like Cato, Giacomelli, Kotler, Funk. These directly measure the protein's activity level. Super important data.
[00:05:49] Speaker B: Okay, functional assays, then what?
[00:05:52] Speaker A: Then they layered on computational predictions. Basically, how likely do computer algorithms think this variant is to be damaging? They use standard tools like Bazdel, agvgd, and also a really powerful newer one called Alphamesense.
[00:06:06] Speaker B: Right, I've heard of that one. And then they added something kind of new to the mix, didn't they? Immunogenicity. That's not standard in VCEAP yet.
[00:06:12] Speaker A: No, it's not. It was quite innovative. They used predicted immune fitness scores. You can sort of think of it as how easily can the immune system see this altered p53 protein.
[00:06:21] Speaker B: Oh, okay.
[00:06:22] Speaker A: A lower immune fitness score suggests the protein might be better at hiding from the immune system. Less likely to get cleared out, which, you know, could increase its potential to drive cancer. It connects the molecular change to a potential clinical outcome.
[00:06:35] Speaker B: Interesting. So function prediction, immunogenicity, what else?
[00:06:39] Speaker A: Frequency. How often does this variant show up in the general population? They pulled data from gnomed, the big public database.
[00:06:46] Speaker B: Standard practice.
[00:06:47] Speaker A: Yes, but then crucially, they linked all this to real patient clinical data. They had access to a huge dataset from ambry genetics.
Over 256,000 controls compared against 708 people confirmed to carry TP53 variants.
[00:07:03] Speaker B: And that group included both RP and p carriers.
[00:07:06] Speaker A: Exactly. 203 classified as reduced penetrance and 505 standard pathogenic. This let them directly compare cancer rates and critically, the age when people got their first cancer diagnosis.
[00:07:17] Speaker B: Got it. So tying it all together, they built.
[00:07:19] Speaker A: A machine learning model, a random forest model. Specifically the idea was to feed it all nine types of data, Functional, computational, immune frequency, clinical, and train it to spot the subtle patterns distinguishing benign reduced penetrance and pathogenic variants.
[00:07:33] Speaker B: Okay, so what did they find? Did the RP variants just look like slightly weaker P variants?
[00:07:37] Speaker A: That's the fascinating part.
No, they didn't just cluster near the pathogenic end. They consistently carved out their own distinct intermediate space across almost every single measure they looked at.
[00:07:49] Speaker B: Really? Like clearly in the middle.
[00:07:51] Speaker A: Clearly in the middle. Take the functional scores. The median scores for the RPE variants consistently landed smack dab between the benign medians and the pathogenic medians. It wasn't just close, it was intermediate. Wow.
[00:08:05] Speaker B: Any specific assay show that? Really?
[00:08:08] Speaker A: Clearly, the CATO assay was almost like a proof of concept. It's one of the functional assays that actually defines an intermediate or partial function range in its output. Okay, and guess What? A whopping 70% of the RPMishense variants they studied fell squarely into that defined partial function category.
[00:08:24] Speaker B: 70%. Okay, that's convincing.
[00:08:25] Speaker A: And the functional clustering data backed it up. Pathogenic variance, overwhelmingly. Class A, the lowest function group. Benign variance, mostly. Class D, the highest function. And the RP is dominated by class C. Right in between. You see, if your system only looks for a simple wassel function cutoff, you completely miss class C. Which is exactly.
[00:08:44] Speaker B: Why the old system struggled. It lacked that granularity.
[00:08:47] Speaker A: Precisely. We needed that finer detail.
[00:08:50] Speaker B: And did this intermediate pattern hold up with the computer predictions, too?
[00:08:54] Speaker A: It did. Tools like alpha Missense predicted the RP variants as deleterious.
So not benign. But their median scores, the actual numbers, were consistently lower than the scores for the standard pathogenic variants.
[00:09:07] Speaker B: Like deleterious late?
[00:09:08] Speaker A: Sort of, yeah. And interestingly, that new measure, the immune fitness prediction, it also put the RP variants in an intermediate position, suggesting maybe a slightly lower ability to dodge the immune system compared to the most dangerous pathogenic variants.
[00:09:23] Speaker B: Okay, now, what about frequency? You said they looked at Gnome add D data, right?
[00:09:27] Speaker A: And here's where it could get tricky. For a clinical lab, just using standard filters, the RP variants actually had a significantly higher allele frequency, the population, compared to the standard pathogenic variants.
[00:09:38] Speaker B: So they looked more like benign variants in that sense.
[00:09:40] Speaker A: On the surface, yes.
Which is a critical point. How do you tell them, apart from genuinely common, harmless variants? Yeah, well, the researchers showed that even though they were more frequent than the classic pathogenic ones, most RP variants still met the clinical criteria, like the PM2 supporting rule that argue for pathogenicity.
[00:10:00] Speaker B: Meaning they're still too rare to be considered truly benign.
[00:10:03] Speaker A: Exactly. But maybe common enough to suggest they aren't usually lethal very early in development or in utero. It's a delicate balance.
[00:10:11] Speaker B: Okay, but let's get to the bottom line. For patients, the clinical data, you said it was the headline.
[00:10:16] Speaker A: Oh, absolutely. This is where the difference really hits home. The cancer presentation was starkly different for standard pathogenic carriers. The average age at their first cancer diagnosis was about 36, and a half years.
[00:10:28] Speaker B: Pretty young.
[00:10:29] Speaker A: Very young. Now for the reduced penetrance carriers, that average age jumped significantly. It was 47.7 years.
[00:10:36] Speaker B: Wow. That's more than a decade later.
[00:10:39] Speaker A: Over a decade difference in average onset time. That changes the entire conversation about when and how intensely you need to start surveillance, doesn't it?
[00:10:46] Speaker B: It absolutely does. Can we quantify that risk difference a bit more? Maybe with odds ratios?
[00:10:51] Speaker A: Yes. They did that for early breast cancer diagnosed before age 31. The standard pathogenic variants showed a huge enrichment. An odds ratio of over 16 compared to controls.
Massive risk.
[00:11:04] Speaker B: 16 times the risk.
[00:11:05] Speaker A: Okay.
[00:11:06] Speaker B: And the RP variance?
[00:11:06] Speaker A: They also showed a significant enrichment. Definitely not benign. But the or was much lower. 3.05.
[00:11:13] Speaker B: Okay, so still triple the risk. Which is serious.
[00:11:16] Speaker A: Absolutely. Serious requires management. But it's more than five times lower than the risk in the standard high risk LFS group.
[00:11:22] Speaker B: That completely justifies a different approach to screening and management.
[00:11:25] Speaker A: It confirms it. Yeah. Yeah. It shows these RP variants cause a distinctly attenuated LFS phenotype, milder, later onset.
[00:11:34] Speaker B: So the model they built using all nine factors, could it actually pick out these RP variants?
[00:11:38] Speaker A: It performed really well.
First they tested it on the known variants, the BMP groups they started with. And it classified most of them correctly. Good sign.
[00:11:45] Speaker B: Okay.
[00:11:46] Speaker A: Then they threw in a real challenge. 590 TP 53 variants currently sitting in Clinvar listed as uncertain or having conflicting classifications.
[00:11:54] Speaker B: The VUS pile?
[00:11:55] Speaker A: Pretty much.
And the model went through them and flagged 106 additional missense variants that showed molecular and computational features strongly matching that reduced penetrance profile.
[00:12:06] Speaker B: They'd defined 106 potential RP variants pulled out of the VUS category. That's huge.
[00:12:11] Speaker A: It's a really significant step. It could help clear up a big chunk of that uncertainty backlog, at least for TP53.
[00:12:17] Speaker B: This whole study feels like it fundamentally shifts how we should think about variant classification.
[00:12:23] Speaker A: I think it does. It proves that the tools we have, the functional assays, the prediction algorithms, they can distinguish these RP variants. We just need to use them with more nuance, less reliance on simplistic cutoffs.
[00:12:35] Speaker B: So it's a call to refine the VCEP guidance?
[00:12:38] Speaker A: Absolutely. A direct call. We need to stop treating functional data as just pass fail. We should be using the full resolution these assays offer, especially ones like Kato, that give us intermediate categories like class A, B, C, D, that improve specificity for RP variants.
[00:12:53] Speaker B: And the prediction tools need an update too.
[00:12:55] Speaker A: They do. Right now they mostly spit out deleterious or neutral, they need recalibrating. We need them to explicitly define and score an intermediate category using their continuous scores to separate RP from standard P.
[00:13:08] Speaker B: It also makes you rethink the clinical evidence. Right?
[00:13:12] Speaker A: That's a crucial point. The study strongly suggests that clinical criteria need to adapt. We need to formally recognize attenuated LFS.
[00:13:20] Speaker B: Phenotypes, meaning a later age of onset or lower. Cancer enrichment shouldn't automatically weaken the case for pathogenicity.
[00:13:27] Speaker A: Exactly. Instead, it should be seen as evidence supporting a different type of pathogenicity, reduced penetrance pathogenicity. It moves us away from this one size fits all diagnosis for lfs.
[00:13:37] Speaker B: Now we should probably mention the limitations like the authors did.
[00:13:40] Speaker A: Of course, there's always a potential for some circularity. Maybe some variants got labeled RP initially because of early clinical observations.
[00:13:47] Speaker B: Right.
[00:13:48] Speaker A: And importantly, the number of RP variants used to actually train the model was relatively small. So those 106 newly identified candidates, they definitely need further validation in larger studies.
[00:13:59] Speaker B: Absolutely essential.
[00:14:00] Speaker A: Next step, we need those bigger studies to confirm these candidates and really critically, to nail down the precise age specific cancer risks associated with them. That's how we move from a predictive model to solid clinical guidelines.
[00:14:15] Speaker B: So what does this all mean?
[00:14:17] Speaker A: Well, I think the study convincingly shows that reduced penetrance TP53 variants are a distinct intermediate class. You can see it in their molecular function, their computational scores, their clinical effects. By refining our classification models, by getting them to recognize this middle ground, the whole genomics community can start acknowledging these variant level differences. And that directly informs how we manage patients.
[00:14:41] Speaker B: Leading to more tailored personalized screening strategies.
[00:14:44] Speaker A: Exactly. Customized plans that avoid both dangerous underscreening and potentially unnecessary or overly intensive surveillance. It's about finding the right level for the specific variant's risk.
[00:14:55] Speaker B: Which leads to the big question, right?
[00:14:57] Speaker A: What does this mean for personalized cancer surveillance protocols globally, not just for life from any. But as we inevitably start finding similar reduced penetrance variants in other major cancer genes. Think BRCA1PTN. How do we adapt our entire approach?
[00:15:13] Speaker B: That is a fantastic question to leave everyone thinking about how precision genomics is evolving beyond just finding a mutation to understanding the precise risk level tied to it. Really fascinating stuff.
[00:15:24] Speaker A: This episode was based on an open Access article under the CC BY 4.0 license. You can find a direct link to the paper and the license in our episode description. If 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. Thanks for listening and join us next time as we explore more Science Base by days.
