Researchers at the University of Hong Kong have built an AI tool that can flag your risk of heart disease up to 15 years before symptoms would normally show up — using a single blood draw instead of the genetic tests or invasive screening that risk prediction has traditionally relied on. The tool, called CardiOmicScore, was published in the journal Nature Communications, and it’s already outperforming standard genetic risk scores in the data researchers have tested it on.
What It Actually Does
CardiOmicScore was built by a team from the LKS Faculty of Medicine at the University of Hong Kong, led by Professor Zhang Qingpeng, using deep learning trained on data from the UK Biobank — specifically, nearly 2,920 circulating proteins and 168 metabolites drawn from participants’ blood, combined with genomic data. Rather than looking at your DNA alone, which tells you a fixed, lifelong risk profile, the model reads your current biological state — the actual proteins and metabolic markers circulating in your blood right now. As the research team put it, your genes set the stage, but your proteins tell today’s story.
That distinction matters practically: a genetic test gives you the same answer at 25 and at 55, but CardiOmicScore’s readings shift as your actual health does, meaning it could theoretically be re-run periodically to track whether lifestyle changes are moving the needle on real risk, not just showing a number that was fixed at birth. In the published data, the model didn’t just detect risk for a single condition — it estimated risk across six different cardiovascular conditions from the same blood draw.
Why This Isn’t Available to You Yet
This is the part worth being very clear about: CardiOmicScore is a research finding, not a product. It hasn’t been approved by the FDA, it isn’t available through any doctor’s office or lab, and it hasn’t been validated in a prospective clinical trial. The data behind it comes from the UK Biobank, a population that skews heavily white and British — which means the model’s accuracy for other populations genuinely hasn’t been established yet, and the researchers themselves flag this as a limitation that needs more work before the tool could be used broadly.
In plain terms: this is a real, peer-reviewed, promising result, not a test you can ask your doctor for next week. Nothing here should be treated as medical advice or a substitute for actual cardiovascular screening — if you’re concerned about heart disease risk, that conversation belongs with your doctor using tools that exist today, not a research tool still years from clinical use.
How It Compares to What Your Doctor Already Checks
A standard cholesterol panel measures a handful of known markers — LDL, HDL, triglycerides — that have decades of clinical validation behind them but paint a fairly narrow picture. Polygenic risk scores, the genetic alternative that’s become more common in recent years, look at inherited risk based on your DNA but can’t account for anything that’s happened to your body since birth: diet, exercise, existing inflammation, or medication. CardiOmicScore’s pitch is combining the breadth of nearly 3,000 protein and metabolite measurements with genomic data in a single model, which is a meaningfully larger dataset than either approach uses alone — that’s part of why the published results showed it outperforming polygenic scores specifically, even though it’s nowhere near ready to replace what your doctor currently uses.
What’s Next
The realistic path from here is years, not months: validation in more diverse populations, a prospective clinical trial to confirm the retrospective findings hold up moving forward rather than just backward through existing data, and eventually — if all of that pans out — some kind of regulatory approval process before it could show up as an actual clinical test. The research team has signaled interest in expanding validation, but hasn’t published a specific timeline for when, or if, this becomes something you can actually take.
If wearable health tracking is more your speed while you wait for research like this to mature, we’ve covered how to pick a fitness tracker that doesn’t require a subscription, and our review of the Google Fitbit Air digs into what today’s consumer health hardware can and can’t actually tell you about your heart health right now.
Deixe um comentário