0
0
In a groundbreaking study, researchers from Penn State College of Medicine have developed an artificial intelligence (AI) tool capable of predicting the progression of autoimmune diseases from preclinical stages. These diseases, in which the immune system mistakenly attacks healthy cells and tissues, often present mild symptoms or specific antibodies in the blood before advancing to full disease stages. Early detection and intervention could significantly improve treatment outcomes, but identifying individuals likely to progress remains a challenge.
The new method, published in Nature Communications, leverages AI to analyze electronic health records (EHRs) and genetic data, producing a risk prediction score for disease progression. This “Genetic Progression Score” (GPS) is reportedly 25% to 1,000% more accurate than existing models in predicting who will advance to later disease stages.
A Targeted Approach to Prediction
“By focusing on individuals with early symptoms or a family history of autoimmune disease, we can use machine learning to identify those at highest risk and tailor interventions accordingly,” said Dajiang Liu, a distinguished professor at Penn State and co-lead author of the study. Liu emphasized the importance of early detection, as autoimmune diseases often cause irreversible damage once they progress.
Using transfer learning, a machine learning technique that adapts pre-trained models to new tasks, the team created GPS to work with limited data. This approach combines insights from genome-wide association studies (GWAS) and EHR-based biobanks to enhance predictive accuracy.
Data-Driven Insights
GPS integrates large-scale genetic data from GWAS to identify genetic differences linked to autoimmune diseases. This information is combined with EHR data, which provides detailed clinical information, such as genetic variants, lab results, and diagnoses. By merging these datasets, the researchers refined the GPS model to predict disease progression with unprecedented accuracy.
The study tested GPS on data from the Vanderbilt University biobank for conditions like rheumatoid arthritis and lupus and validated its performance using the National Institutes of Health’s All of Us biobank. GPS outperformed 20 other models in predicting disease progression, underscoring its potential to revolutionize early diagnosis and treatment.
Transforming Patient Care
“This innovation could change the way we approach autoimmune diseases,” said Bibo Jiang, lead author and assistant professor at Penn State College of Medicine. “By identifying individuals at high risk, we can enable targeted monitoring, early interventions, and personalized treatment plans.”
Liu added that the GPS model could also improve clinical trial design by identifying patients most likely to benefit from experimental therapies. Beyond autoimmune diseases, the framework could be adapted to study other conditions, potentially reducing health disparities in underrepresented populations.
A Promising Future
Approximately 8% of Americans, predominantly women, live with autoimmune diseases. Early interventions could significantly alter their disease trajectories, making predictive tools like GPS invaluable. By integrating cutting-edge AI with comprehensive data, this research highlights the transformative potential of technology in medicine.
The findings mark a significant step forward in precision medicine, offering hope for millions affected by autoimmune diseases worldwide.
For more details, refer to the study: Chen Wang et al, “Integrating electronic health records and GWAS summary statistics to predict the progression of autoimmune diseases from preclinical stages,” published in Nature Communications (2025).
