Blood Doesn't Speak for Us
When autism is reduced to a blood signature, the child disappears
The newest entrant in autism biomarker research comes from a Korean research team working out of Asan Medical Center. Their paper, currently in press at Progress in Neuropsychopharmacology & Biological Psychiatry, claims that seven proteins in dried blood spots can differentiate autistic children from non-autistic peers with 96.7% sensitivity. It's a slick, technically impressive study. It's also a structural threat (though early-stage and flagged by the authors as preliminary).
Let’s name it clearly: this is not discovery — it’s screening dressed in science. The language is diagnostic, the frame is preemptive and the logic is surveillance. Blood samples are easier to collect than full behavioral evaluations, especially in overstretched systems. So when the authors cite cost and time as barriers to "gold standard" diagnosis, what they’re offering — what they frame as augmentation — still functions as a shortcut in logic because it translates personhood into a protein marker, and calling it precision.
Proteomic studies don’t have to be harmful. But this one is. Not because its methods are invalid, but because its framing is. The children in the study aren't treated as people with divergent experiences. They’re reduced to signal — not understood, just sorted. Correlations are drawn between protein levels and standardized checklists like the ABC and CARS, and the fact that autistic traits map inversely onto protein levels is treated not as a bias in measurement, but as validation of the marker. If that logic holds, then the most "autistic" children are simply those with the most deviant biochemistry. To be clear, the authors present this as correlation, not defect — but the framing risk lies in what that correlation enables: a logic where autistic behavior is read through deviation.
Nowhere in this paper do we hear what these traits mean. What they feel like. How they are shaped by environment, by interaction, by interpretation. The children are not asked about their own social landscapes. The blood speaks; the autistic child doesn't get to.
This matters because the stakes are not academic. Biomarkers are not neutral. They shape who gets flagged — and who gets fixed. When you label something as a preclinical signal, you invite early intervention — often without consent, often before a child can communicate, often in ways that pathologize difference as delay. The authors know this. They write that early detection is critical, that intervention changes outcomes, that cost-effective tools are needed. But cost-effective for whom — the children and their families, or the system that wants to reshape them?
This is not just a proteomics paper. It is an ideological one. It operates on the premise that the diagnostic bottleneck is the biggest problem in autism. But if you're autistic, that bottleneck might be the last defense between you and a system eager to change who you are before you've even had a chance to know it. If this model goes to scale, autistic children — especially those with intellectual disabilities — could be flagged and sorted long before they understand what that means. And once flagged, the burden of proof will shift. The assumption will be that intervention is needed. That the brain is broken. That the proteins have spoken — and the child is already rewritten.
We refuse that logic. Proteins are not pathology. Blood does not consent — especially when it’s collected to preempt a voice that hasn’t spoken yet. And autistic children are not data points in a diagnostic arms race.
If there is science worth funding, it is science that starts with autistic insight, not molecular shortcuts or smears of blood. It would ask different questions, use different measures, and build models with autistic people instead of extracting them. It might begin not in the lab but in a classroom, a family, a moment of overload that demands understanding instead of calibration. It is science that asks why some children shut down in classrooms, or go nonverbal in hospitals, or refuse to make eye contact in tests because they’ve learned that being seen often means being corrected. It is science that knows the difference between signal and symptom, and refuses to collapse the two.
The authors are clear that the study is preliminary, with small sample size, overfitting risk and no external validation — but those brakes don’t stop the framing from circulating.
The authors call for a larger cohort to replicate their findings. We call for a deeper question: What happens when you get what you want? When the biomarker works, when the protein fingerprint becomes a diagnostic tool? What world gets built for the child flagged by a lab slip — before they can speak for themselves?
We don't need faster detection. We need a better frame. And no blood test can give us that.
Note: This critique targets the framing logic embedded in the study — not the intent of its authors. The paper itself is cautious: it names the risk of overfitting, acknowledges the small and narrow sample, and calls for replication before any clinical use. It presents protein correlations as candidate markers, not replacement diagnostics. Still, framing matters — and framing scales faster than science.