New AI Tool Revolutionises Drug Discovery for Rare Diseases
Harvard Medical School, 25 September 2023 — In a promising development for the global healthcare landscape, scientists at Harvard Medical School have unveiled a groundbreaking artificial intelligence (AI) model designed to expedite the discovery of treatments for rare and undiagnosed diseases. This new tool, named TxGNN, is the first AI system specifically developed to identify drug candidates for conditions with limited to no existing therapies.
An Expansive Health Crisis
There are currently over 7,000 rare and undiagnosed diseases worldwide, affecting approximately 300 million people. Despite the collective scale, only about 5 to 7% of these diseases have an FDA-approved drug to manage them, leaving a majority of patients with inadequate treatment options. The introduction of TxGNN aims to bridge this gap, offering hope to millions and potentially reducing the human and economic toll these conditions impose globally.
Unprecedented Scope and Results
TxGNN leverages existing data from thousands of medicines to identify new therapeutic candidates. Remarkably, during its initial experiments, the AI tool identified drug candidates for more than 17,000 diseases, including many with no current treatments, offering an unprecedented scope of application. Researchers from Harvard made the tool publicly available in an effort to encourage its use among clinician-scientists focusing on neglected conditions.
"With this tool, we aim to identify new therapies across the disease spectrum. For rare, ultrarare, and neglected conditions, we foresee this model could help narrow a gap that creates serious health disparities," said Marinka Zitnik, assistant professor of biomedical informatics in the Blavatnik Institute at Harvard Medical School and lead researcher of the project.
Efficient and Cost-Effective Alternative
One of the prime advantages of TxGNN is its ability to repurpose existing drugs—both FDA-approved and those currently undergoing clinical trials. This approach is not only faster but also more cost-effective than developing new drugs from scratch. It capitalises on the well-understood safety profiles and known effects of existing medications, making the drug discovery process less reliant on serendipity and more on strategic data analysis.
The tool was able to identify drug candidates using data from nearly 8,000 medicines. Additionally, it predicted potential side effects and contraindications, information typically gathered through early-phase clinical trials. Compared to leading AI models for drug repurposing, TxGNN was about 50% more effective in identifying drug candidates and 35% more accurate in predicting contraindications.
Advanced Analytical Capabilities
Unlike most AI models used for drug discovery that focus on a single disease, TxGNN can identify shared features across multiple conditions. This ability enables it to draw parallels between well-understood diseases and those less understood, offering the potential for novel treatment options. The AI was trained on vast datasets, including DNA information, cell signalling pathways, levels of gene activity, and clinical notes.
The researchers validated the tool’s efficacy using 1.2 million patient records, and it successfully identified drug candidates for various diseases. Notably, the tool explained its reasoning in a manner consistent with existing medical knowledge, increasing transparency and physician confidence.
Clinical Applications and Future Outlook
TxGNN's potential applications extend beyond rare diseases. "New drugs could offer alternatives with fewer side effects or replace drugs that are ineffective for certain patients," Marinka Zitnik noted, highlighting the broader implications for more common health conditions.
The Harvard team cautions that any therapies identified will require further evaluation for appropriate dosing and administration timing. However, with its unique analytical strength, TxGNN is set to expedite the drug repurposing landscape, providing a strategic, data-driven approach hitherto unachievable.
The team has already started collaborations with several foundations focusing on rare diseases to identify possible treatments using this AI tool, marking an exciting juncture in the fight against global health disparities.
Publication and Further Information
The detailed findings and methodology behind TxGNN are documented in the latest issue of Nature Medicine.
Source: Noah Wire Services