How do you advance disease research when the patient data needed to power it is difficult to access?
For life sciences organizations, the answer to this question is rarely straightforward. Understanding how diseases evolve over time requires longitudinal clinical records, genomic information, therapeutic histories and patient outcomes spanning multiple years. Yet access to much of this data is limited by privacy requirements, governance restrictions, access controls and technical limitations that make it difficult to use in research environments.
For a global pharmaceutical company, these constraints became a significant obstacle in advancing obesity research. The organization sought to build an AI-powered model to study how obesity progresses across different patient populations. The goal was to deepen clinical insights and support future research and therapeutic development.
However, access to longitudinal clinical, genomic and therapeutic data proved difficult. Potential sources such as UK Biobank came with restrictions around access, data portability and usage. Critical multi-omics and genomic datasets could not easily be moved into the environment required for model development. Even where access pathways existed, approval timelines did not align with the need to rapidly demonstrate the feasibility of the approach.
As the initiative progressed, it became clear that solving the data challenge would be just as important as developing the model itself.
Creating a Usable Foundation for Research
Rather than allowing data constraints to slow the program, Persistent proposed a different path.
Working with domain experts and reference datasets, the team developed a synthetic data strategy that could replicate realistic patient journeys while respecting governance and data-access requirements. Instead of waiting for access to restricted datasets, researchers could work with a representative longitudinal dataset designed specifically for the program.
The resulting dataset represented approximately 2,500 patients over a five-year period and generated roughly 10,000 patient records capturing disease progression patterns, clinical events and patient trajectories. The client’s domain experts reviewed and validated the output, providing confidence that the synthetic data reflected realistic disease characteristics and progression patterns.
What started as a response to a data-access challenge quickly became a strategic advantage. Research teams were no longer constrained by lengthy approval cycles, limitations around data movement or uncertainty around the availability of longitudinal patient records. They could begin exploring disease progression patterns and patient cohorts while remaining aligned with privacy and governance requirements.
Turning Synthetic Data into Disease Intelligence
With a usable data foundation in place, the focus shifted from obtaining data to understanding disease progression.
Persistent fine-tuned foundation models, including Google’s MedGamma to incorporate clinical, genomic and therapeutic relationships relevant to obesity research. This enabled the solution to analyze connections across patient characteristics, treatment pathways, molecular signals and disease outcomes rather than relying on a single source of information.
The resulting framework supported:
- Disease progression analysis
- Patient stratification
- Risk assessment
- Phenotype and endotype identification
- Precision medicine exploration
- Therapeutic target discovery
- Clinical interpretation support
Researchers gained a longitudinal view of obesity progression across patient populations over time. Instead of working with fragmented information, they could study progression pathways, identify clinically meaningful patient subgroups and explore how different factors influenced disease development.
The project ultimately produced obesity progression models covering approximately 2,500 patients and 10,000 records across five years of simulated patient histories. The capability met the objectives established for the program and demonstrated that meaningful disease progression modeling could be achieved despite significant data-access constraints.
Creating a Framework Beyond Obesity Research
While obesity served as the initial focus, the broader opportunity extended well beyond a single disease area.
The project built upon earlier work on the progression of insulin resistance and demonstrated a repeatable framework that can be adapted to other disease progression initiatives. The same approach can support future research involving metabolic disorders and therapeutic areas where longitudinal patient data remains difficult to obtain, use or share directly.
This expanded the value of the initiative significantly. Rather than creating a model for one research question, the organization established a repeatable method for advancing disease research when traditional data-access pathways prove restrictive, incomplete or slow to support innovation.
Advancing Research Within the Realities of Healthcare Data
The most important outcome was not a disease progression model or a synthetic dataset.
It was demonstrating a practical way for research to move forward when access to real-world patient data is constrained.
By combining synthetic data generation, domain expertise and foundation-model fine-tuning, Persistent helped researchers shift their focus from overcoming data-access challenges to understanding disease progression. The team successfully modeled obesity progression across approximately 2,500 patients and 10,000 records spanning five years, creating a foundation for patient stratification, risk assessment, therapeutic exploration and future precision medicine initiatives.
The implications extend beyond obesity. The same approach can be applied to other metabolic conditions and disease progression programs where longitudinal patient data remains difficult to obtain, use or share. The framework established a repeatable path for advancing research while operating within the realities of modern healthcare governance.
As life sciences organizations pursue increasingly ambitious questions around disease progression, patient outcomes and personalized therapies, access to data will remain one of the industry’s defining challenges. This initiative demonstrated that those challenges do not have to limit innovation.
With the right combination of synthetic data, AI and scientific expertise, organizations can continue exploring complex diseases, generating new insights and accelerating discovery. Because sometimes the next breakthrough begins not with more data, but with a better way to learn from the data that is already within reach.




