Why AI Can’t Fix Healthcare Until It Understands Biology

AI promised customized care, but it surely nonetheless treats sufferers as knowledge. Actual progress begins when it could actually actually perceive the physique itself.

getty

Within the final decade, “personalization” has grow to be one of the crucial overused phrases in medication. However even with the various guarantees to tailor take care of every affected person, hundreds of thousands nonetheless shuffle via the identical diagnostic routines, prescriptions and unwanted side effects.

It’s the paradox shaping a lot of AI-powered healthcare at this time. Whereas synthetic intelligence has made some main elements of healthcare extra environment friendly, it’s not essentially made it extra human. It’s like an phantasm of particular person care that’s left many sufferers feeling extra like knowledge factors than folks.

The issue, consultants say, isn’t the expertise itself however the form of knowledge we feed it. Most AI fashions in healthcare at this time are educated on population-level datasets resembling digital well being information and claims data. Whereas these datasets can reveal statistical tendencies throughout hundreds of thousands of sufferers, they not often seize what’s actually occurring inside any one in every of their our bodies. In different phrases, they will predict chances, however not organic realities.

As Mika Newton, CEO of xCures, said earlier this 12 months, “AI won’t remodel healthcare if it operates in a vacuum. AI requires the inspiration of high-quality knowledge, which begins with sufferers. That personalization drawback is what startups like California-based Parallel Health are actually attempting to unravel by serving to AI interpret organic knowledge immediately.

From Knowledge Factors To Residing Methods

“Actual personalization means treating you as a posh system, not a statistic,” mentioned Natalise Kalea Robinson, cofounder and CEO of Parallel Well being. “Most ‘customized’ healthcare at this time is de facto simply refined segmentation — you’re positioned in a bucket based mostly on signs — generally demographics or genetic markers (for those who’re fortunate), then given the therapy that labored for most individuals in that bucket.”

Parallel’s work is one instance of how corporations are starting to make use of biology as the inspiration for personalization quite than relying solely on medical information or demographics. Its platform makes use of quantitative whole-genome sequencing to map the trillions of micro organism, viruses, and fungi that make up an individual’s pores and skin microbiome. “This isn’t about evaluating you to a inhabitants common — it’s about understanding your particular person organic actuality on the microbial degree and at a number of factors,” Robinson defined.

Two sufferers might share an pimples prognosis, however their underlying causes may be completely totally different. One affected person may need an overgrowth of Cutibacterium acnes phylotype 1A, the micro organism often linked to pimples, whereas one other may need antibiotic-resistant sorts that specify why common remedies didn’t work. “No two ‘pimples’ sufferers have the identical pores and skin microbiome; we have now but to see that throughout our extremely giant knowledge set,” Robinson mentioned.

That organic specificity permits the corporate to design focused phage serums that remove solely dangerous strains whereas preserving helpful microbes. Exterior researchers agree that such a precision is scientifically promising, although they notice that phage therapy nonetheless faces steep regulatory, manufacturing and standardization hurdles earlier than widespread adoption.

Robinson acknowledges these challenges however argues that adaptability — not simply precision — will decide which approaches endure. “Your biology isn’t static, so your therapy shouldn’t be both,” she mentioned. “Actual personalization is longitudinal, adaptive, and grounded in your precise organic knowledge — not inhabitants proxies.”

Instructing AI To Perceive Trigger And Impact

For years, healthcare AI has been praised for sample recognition — recognizing tumors in scans, predicting readmissions and flagging anomalies in lab outcomes. However Dr. Nathan Brown, Parallel Well being’s chief science officer, argues that’s solely the floor. “Working with direct organic knowledge transforms AI from a pattern-matching software right into a mechanistic prediction engine,” he mentioned.

By analyzing how microbes work together with each other and with the human host, the system can start to deduce causality quite than mere correlation. “Our AI can establish that particular microbial imbalances preceded symptom onset by months, enabling true prediction, not simply early detection,” Brown famous.

That perception, he mentioned, turns AI from reactive to preventive medication. The identical microbial patterns that sign irritation in pimples, for example, may seem in situations like rosacea or sure varieties of psoriasis. “What we study microbial dysbiosis in a single situation can apply to others. Our AI is studying basic ideas of host-microbe interplay that generalize throughout illnesses. We then have the facility to redefine advanced illnesses.”

Whereas unbiased researchers have echoed the potential of biology-driven AI methods, particularly these based mostly on microbiome knowledge, they continue to be cautious, as famous in a review printed in Nature.

Scaling The Science

The phrase “customized” usually evokes hand-crafted medication — remedies so particular they will’t presumably scale. Dr. Seaver Quickly, Parallel’s lead dermatologist and scientific advisor, mentioned that assumption misses how platform applied sciences evolve.

“Personalization doesn’t imply we’re creating distinctive remedies for each particular person from scratch,” he mentioned. “We’re utilizing platform expertise to effectively match sufferers to a bespoke resolution from an outlined toolkit.” Parallel claims its ‘toolkit’ attracts on an increasing biobank of microbial strains and a producing course of aimed toward stabilizing focused phage therapies — a problem the broader biomanufacturing area can be racing to unravel.

That mannequin mirrors the early days of genomic medication, when sequencing DNA was gradual and costly however finally turned routine. The identical may occur with microbiome-based care because the expertise matures. “Precision medication eliminates trial and error,” Robinson defined. “If we are able to inform from the beginning {that a} affected person’s micro organism are immune to sure antibiotics, we are able to keep away from remedies that gained’t work, saving each time and value.”

Current analysis helps that concept, with a review within the Journal of Translational Medication noting that whereas precision therapies can enhance outcomes and scale back waste, their cost-effectiveness nonetheless depends upon reimbursement insurance policies and entry — two long-standing obstacles to progress in scientific genomics.

The Moral Edge

As biology-driven AI turns into extra highly effective, questions relating to privateness and fairness have gotten more and more outstanding. A report from the Nationwide Middle for Biotechnology Info warned that “using giant datasets in AI methods has led to discussions about possession and administration of knowledge,” including that knowledge sovereignty — the suitable of people or teams to regulate how their organic knowledge is collected and interpreted — will outline the subsequent section of well being innovation.

In accordance with Robinson, that precept is already constructed into Parallel’s mannequin. “Sufferers should know what knowledge is collected, how it will likely be used and what they get in return,” she mentioned. “Simply because you may gather organic knowledge doesn’t imply you must.”

She believes that transparency and equitable entry should coexist. “Probably the most harmful danger in customized medication is making a two-tier system the place precision care is obtainable solely to the rich. “Communities that contribute knowledge to our AI fashions should profit from the ensuing enhancements.”

Bioethicists are more and more voicing related issues. Current analysis — together with a 2024 paper in BMC Medical Ethics by Shaw and colleagues and a 2025 study printed by the Committee on Knowledge for Science and Expertise — emphasizes that the way forward for customized medication will rely not solely on smarter algorithms however on fairer methods of belief, consent and shared profit.

Robinson calls it a shift of energy again to the affected person — a much-needed correction at a time when knowledge privateness stays a defining concern. Whether or not healthcare follows that path will rely on AI’s potential to account for the organic complexity of every particular person, quite than simply patterns in inhabitants knowledge.