Slashing adverse drug reactions through pharmacogenetics and advanced AI could help rehabilitate the pharmaceutical industry’s reputation amid mounting criticism.
The pharmaceutical industry is facing an unprecedented crisis of trust. Beyond the well-publicized issues of soaring drug prices, the opioid epidemic and aggressive marketing tactics, a more insidious problem persists: the very medications intended to heal are, in some cases, causing significant harm.
Adverse drug reactions (ADRs) now rank among the top five causes of death in the U.S., contributing to as many as 175,000 fatalities each year and imposing a financial burden of more than $150 billion, according to FDA data. An estimated 90% of these events go unreported, leaving much of the human toll hidden from view.
At the core of this crisis lies a fundamental flaw in modern medicine: a one-size-fits-all approach to prescribing. Physicians still rely heavily on trial and error to determine which medications and dosages will work best for individual patients. This imprecision not only leads to treatment failures and dangerous side effects but deepens a growing skepticism about the integrity and effectiveness of the pharmaceutical industry.
That erosion of trust is no longer confined to patients and advocacy groups. Recent leadership changes across the Department of Health and Human Services reflect shifting official perspectives on the pharmaceutical industry. HHS Secretary Robert F. Kennedy Jr.’s “Make America Healthy Again” initiative emphasizes reducing what he characterizes as overprescription of antidepressants and raises concerns about vaccine safety testing protocols. Marty Makary, who now leads the FDA, has expressed concerns about “an epidemic of inappropriate care” and “massive overtreatment,” particularly regarding antibiotics, in his 2024 book Blind Spots: When Medicine Gets It Wrong, and What It Means for Our Health. Centers for Medicare & Medicaid Services Administrator Mehmet Oz has been noted for his support of alternative therapies, which has drawn criticism from some in the medical community. Together, these developments suggest a shift in how healthcare practices are evaluated.
For decades, pharmacogenetics—the study of how genes influence drug response—has been seen as a way to make prescriptions more precise and, therefore, safer and more effective. The problem, however, has always been scale. Drug response is not determined by single genes alone but by complex interactions between multiple genetic markers, metabolic pathways, environmental influences and lifestyle factors. Manually analyzing these layers of variables to create truly personalized prescriptions is beyond human capability. Even traditional computational models have struggled to make sense of the sheer volume of data.
Now, artificial intelligence is changing that. As co-founder of the pharmacogenetics startup PGxAI, I have a front-row seat to how advanced AI platforms can integrate vast, multilayered datasets encompassing genomics, transcriptomics, epigenetics and even diet and microbiome profiles. These systems can learn from diverse patient populations around the world and adapt to account for differences in drug metabolism. Just as importantly, they improve over time as they ingest real-world outcomes from patients who follow pharmacogenetic guidance.
Big Data Meets Big Processing Power
The potential impact of these systems is profound. Studies indicate that 98% of individuals carry at least one genetic variant that could affect how they metabolize commonly prescribed drugs. Meta-analyses suggest that pharmacogenetic insights can reduce the incidence of adverse drug reactions by more than 30%. That’s not a marginal improvement—it’s a direct path to reducing one of the country’s leading causes of preventable death.
Until recently, however, implementation remained elusive. Pharmacogenetics was caught in a classic catch-22: without large-scale, real-world datasets, there was no way to train AI; without capable AI systems, there was little incentive to collect the data.
That is changing. Major health systems including Mount Sinai, Massachusetts General Hospital, Mayo Clinic and the Department of Veterans Affairs are now collecting genomic data alongside clinical records. At Mount Sinai, the Million Health Discoveries Program, launched in partnership with the Regeneron Genetics Center in 2022, seeks to sequence 1 million patients and integrate their genetic data with de-identified electronic health records. The VA’s Million Veteran Program, initiated in 2011, has already surpassed 1 million participants, making it one of the world’s largest and most diverse biobanks. Mayo Clinic’s long-established biobank has extended its national reach through collaboration with the All of Us Research Program.
These initiatives aim to create real-world models for embedding genomics into routine care. Internationally, initiatives like the nationwide Saudi Genome Program, launched in 2018, demonstrate the global momentum toward genomics-driven healthcare.
Concurrently, the capabilities of AI have grown exponentially. Today’s models can process enormous volumes of data in real time, providing clinical insights at the point of care.
Putting Insights Into Practice
The infrastructure is finally in place. Where pharmacogenetic recommendations used to crawl forward, requiring years of validation and consensus through the FDA or expert consortia, AI now enables dynamic, real-time updates grounded in real-world performance. What’s still missing is coordination: a unified system to harness this infrastructure and turn decades of scientific progress into personalized care at scale.
Still, some pharmaceutical companies remain hesitant. Their concern is that pharmacogenetics might narrow their market by limiting broad-based prescriptions. But in cardiology—where genetic testing is already standard practice and widely reimbursed—the benefits are clear. Take warfarin, for example. Patients with certain genetic variants face a dramatically increased risk of serious bleeding if treated with standard doses of the drug. In psychiatry and neurology, where patient responses can vary dramatically, precision tools offer equally transformative potential.
Even blockbuster drugs like GLP-1 agonists carry serious risks when used broadly—including pancreatitis, gallbladder complications, kidney impairment and potential thyroid tumors. Early user data show that nearly one in three patients carry genetic variants linked to increased risk. As these drugs scale, wide adoption must go hand in hand with personalization.
For high-cost therapies, improved precision doesn’t just enhance outcomes; it accelerates adoption, increases efficiency and offers a much-needed path to restoring public trust.
The promise doesn’t end with today’s medications. AI-driven pharmacogenetics could transform the clinical trial process itself, enabling researchers to identify likely responders to experimental therapies with far greater accuracy and increase candidate drugs’ chances of success. It could even help revive drugs previously deemed failures by isolating genetic subgroups that would benefit, turning past losses into future breakthroughs.
Indeed, over the past decade, collection of genetic data during clinical trials has quietly become standard practice across the pharmaceutical industry. Most major companies now routinely obtain DNA samples from trial participants and conduct pharmacogenomic analyses as part of early and late-stage drug development.
What was missing until now was the ability to interpret that data at scale. With AI systems finally able to process it, the industry has reached a turning point in the potential to realize the promise of pharmacogenetics—both in drug development and clinical practice. Whether Big Pharma seizes this opportunity or lets it slip by will define the decade ahead.
