A new bottleneck is emerging in the race to use artificial intelligence for drug development. While tech giants like Amazon, Google and NVIDIA have poured resources into AI systems that can design molecules and predict protein structures, the next critical step requires something far less glamorous: connecting those digital discoveries to real patient records.
Ardy Arianpour, CEO of health data platform SEQSTER, sees this as the missing layer in AI powered pharma. “Drug discovery does not happen in a vacuum,” he said. “It requires real world patient data to contextualize, validate and prioritize candidates.” His company has spent years building a system that aggregates fragmented health records into a single longitudinal view, now covering 158 million de identified patients and 211,000 clinicians nationwide. The platform cleans and standardizes data from electronic health records, labs, claims, genomics and physician notes, making it usable for research.
The value of connected data showed up in a concrete example. When pharmaceutical company AbbVie needed to recruit migraine patients for a clinical trial, SEQSTER identified more than 5,400 eligible candidates within three months. But the analysis also revealed something unexpected: 22 percent of female migraine patients on certain medications had endometriosis noted in their electronic health records. That comorbidity signal, invisible when data stays siloed, could help researchers design entirely new trials or repurpose existing drugs for different conditions.
SEQSTER’s approach builds on a vision Arianpour describes as “the Napster of health data sharing, but in a legal, consented way.” The company started in 2016 by giving patients control over their own records, a concept cardiologist Eric Topol demonstrated in 2018 when he used the platform to assemble his complete medical history. That same infrastructure now serves life sciences customers trying to recruit trial participants, screen cohorts and track real world outcomes beyond scheduled study visits. For example, SEQSTER powers the Multiple Sclerosis Association of America’s registry with Novartis, capturing data when patients visit emergency rooms or start new medications in real time.
Why Chain of Custody Data Matters for AI
As patients begin using general purpose AI tools like ChatGPT to interpret their own medical records, Arianpour warns that self reported information is not reliable enough. SEQSTER has introduced conversational AI for medical records, but it works best when fed “chain of custody” data that comes directly from providers and electronic health record systems. “I would not recommend that someone rely only on their own self reported information in an off the shelf AI tool,” he said.
The next frontier is closing the loop between AI generated molecular hypotheses and the human health data needed to test them. Arianpour believes the technology to solve healthcare’s “real disease of interoperability” is finally catching up with the ambition. As more health systems, labs and imaging centers connect their data, the path from computational discovery to real patient impact becomes clearer. The pieces are coming together, and the patients at the center stand to benefit most.