Jude Walker: Elin, before anything else — I want to give you a number and I want your honest reaction to it.
Elin Cole: Numbers I can do.
Jude Walker: CYP3A4 — one enzyme in the liver — metabolizes approximately 50% of all clinically used drugs. One enzyme. Half the pharmacopoeia.
Elin Cole: That does land differently when you just say it plainly like that.
Jude Walker: And CYP2D6 handles another 25%. So between two Cytochrome P450 enzymes, you're covering three-quarters of what we prescribe. And both of them are subject to genetic polymorphisms — SNPs, copy-number variants — that mean two people on the same drug at the same dose can have completely different blood concentrations. That's not theoretical. That's Clara in the ED while her colleague's fine.
Elin Cole: So the scenario — same ward, same pill, one walks home — that's pharmacogenomic variation made concrete.
Jude Walker: Exactly. And the question I can't shake is: if we've mapped this mechanism, if we know these Cytochrome P450 variants exist and do this — why is fixed dosing still the default? What's actually stuck?
Elin Cole: I think the way to see this is — there's the biological claim, which is solid, and then there's the implementation claim, which is a completely different conversation. And I suspect those are getting collapsed in a way that flatters everyone involved.
Jude Walker: Right — but the collapse you're describing, that's downstream. Before we get there, I want to make sure we actually have the mechanism. Because I don't think most people do.
Elin Cole: Okay, so — think of a CYP enzyme as a dimmer switch on drug clearance. Your genes set where the dimmer sits. And everyone's dimmer is in a different position.
Jude Walker: That's the whole thing, right there.
Elin Cole: And CYP2D6 is the canonical case. Four positions on that dimmer — poor metabolizers, intermediate, extensive, ultrarapid. Poor metabolizers, the drug stacks up, you get toxicity at a dose that's fine for everyone else. Ultrarapid metabolizers, the drug clears so fast it never reaches therapeutic concentration. You've prescribed something that just... isn't doing anything.
Jude Walker: And that's the part that inverts for prodrugs — wait, actually that's the thing most people miss entirely.
Elin Cole: It completely flips the risk direction. Clopidogrel is inactive when you swallow it. CYP2C19 has to convert it to the active form. So a CYP2C19 poor metabolizer isn't overdosing — they're just... not being treated. The active drug never gets made. Same logic with tamoxifen through CYP2D6.
Jude Walker: And it's not just CYP. DPYD variants — severe toxicity from fluoropyrimidine chemotherapy. TPMT and NUDT15, thiopurine toxicity. UGT1A1*28, that's irinotecan — severe neutropenia, severe diarrhea. SLCO1B1, statin-induced muscle damage. Same logic, completely different enzyme families.
Elin Cole: So the dimmer-switch idea scales. It's not a CYP quirk — it's a principle that runs across how we metabolize basically everything.
Jude Walker: Which makes the obvious question: why don't we just test everyone before we prescribe anything?
Elin Cole: Testing everyone assumes the test stays true. That's the crack I want to pull on. Because it doesn't — and that's not a minor caveat.
Jude Walker: Phenoconversion. That's the word. And I think this is the part that breaks the clean story entirely.
Elin Cole: Walk me through it.
Jude Walker: You get genotyped. Extensive metabolizer — CYP2C19 looks fine, dimmer in normal position. Everything checks out. Then six months later your GP adds a strong CYP inhibitor for something completely unrelated — ketoconazole, say, for a fungal infection — and suddenly your actual metabolic capacity collapses. You're functionally a poor metabolizer now. But the genetic test on file still says extensive. It's not wrong, exactly. Your genotype didn't change. Your enzyme activity did.
Elin Cole: So the genotype is stable but the phenotype isn't. The test captured a snapshot of a system that just... keeps moving.
Jude Walker: And enzyme induction works the other direction too — some drugs ramp CYP activity up, independently of what your genes say. Same person, different drug combo, completely different clearance rate. The genotype is basically a baseline that the environment keeps overwriting.
Elin Cole: Right — but I want to push on frequency here. Is phenoconversion a real-world volume problem or more of a theoretical edge case? Because I think the precision promise only actually cracks if this is happening at meaningful scale.
Jude Walker: I mean, CYP3A4 alone metabolizes half of all clinical drugs and it's also the isoform most vulnerable to inhibition and induction from co-prescriptions. So the most common metabolic pathway is also the most unstable one. That's not edge-case territory — that's the center of polypharmacy, which is basically every patient over sixty. Systems pharmacology researchers are actually making this exact argument: pharmacogenomics alone can't explain interindividual variability, you need pharmacometrics, environmental factors, the whole framework. The static test is one input, not the answer.
Elin Cole: So what you actually need alongside any genetic result is real-time drug-interaction awareness — not instead of the test, but the test without that context is genuinely incomplete. And that gets expensive and complicated fast, which is — actually, that connects to something we haven't touched yet: where this testing has actually translated into practice versus where it's just aspirational, which is a messier picture than anyone's advertising.
Jude Walker: And that's exactly the gap. Because the testing that's actually translated — the real wins — it's a short list. DPYD before fluoropyrimidines, TPMT and NUDT15 before thiopurines, UGT1A1 before irinotecan, CYP2C19 and clopidogrel, CYP2D6 and tamoxifen. Those pairs have actual guidelines behind them. But that list is not 'standard practice.' That's like five islands.
Elin Cole: TPMT genotyping before thiopurine initiation — that one actually got MBS funding in Australia. Item 73327. That's a landmark, genuinely. Medicare Benefits Schedule funding means it's reimbursed, it's in the workflow, it's not a flagship-center luxury.
Jude Walker: One line item on a benefits schedule.
Elin Cole: I know, I know — but that's actually the point I want to make. The Human Genome Project finished in April 2003. Three billion dollars, three billion nucleotides sequenced. And that's what unlocked population-scale genotyping — microarrays, next-generation sequencing, the whole infrastructure. So we're now two decades out from that, and the tangible reimbursed output in most health systems is... item 73327 and a few oncology carve-outs. That's the gap I think we need to name directly.
Jude Walker: Three billion dollars and twenty years and most panels are still unfunded. That's not a rollout problem. That's a structural verdict.
Elin Cole: I mean — it splits into two problems, actually. One is clinician education gaps, which are well-documented, prescribers don't know what to do with a pharmacogenomic result even when they have one. The other is workflow integration — how does a genetic result from six months ago surface at the moment of prescribing? Those are solvable, but nobody's solved them. So the biological validity is there, the clinical utility in those five pairs is there, and the implementation is just... not following.
Jude Walker: And the equity consequence of that gap — that's the part nobody's saying. The tests that did get reimbursed, that did get guidelines, those are the ones studied in Western populations. CYP2D6 poor metabolizer hits 10% of Caucasians. Up to 30% of some Chinese populations. The funding follows the evidence base, the evidence base followed the study populations. So the people most likely to need the test are the least likely to be in a system where it's covered.
Elin Cole: So precision medicine built its evidence on the population where the risk was lowest — and called it generalizable. That's not a caveat. That's a design flaw baked into the foundation.
Jude Walker: And that's the thing nobody wants to sit with. The whole promise of pharmacogenomics — we're moving past population averages, we're treating the individual — and the evidence architecture underneath it is built on populations that don't include the people with the highest CYP2D6 poor metabolizer prevalence. Up to 30% in some Chinese populations versus 10% in Caucasians. The guidelines are most precisely tuned for the group least likely to need them most.
Elin Cole: That's not a critique of the science. The molecular mechanism is real. It's a critique of the evidence architecture the science is embedded in — which genomic databases got funded, which populations got enrolled, whose variation got characterized. That's what shaped what became a guideline.
Jude Walker: And closing that gap isn't just — I mean, it's not a matter of running a few more studies. You'd need more diverse genomic databases as the actual foundation, not a supplement. Different funding models. Probably different guideline development processes. That's a rearchitecting, not a patch.
Elin Cole: Which is something I can't quite resolve.
Jude Walker: Nah, me neither. And I think that's actually the honest place — not that the science is wrong, but that the field hasn't caught up to its own premise yet. It says individualization, it's still running on skewed population data. That tension is real and it's open.
Elin Cole: Good conversation. Genuinely — the unresolved ones are always the ones worth having.