AI Model Identifies Potential Treatments for Thousands of Rare Diseases

dots

A new artificial intelligence model developed by scientists at Harvard Medical School offers hope for millions of people worldwide suffering from rare and neglected diseases. The AI tool, named TxGNN, can identify potential treatments from existing drugs for over 17,000 diseases, many of which currently have no approved therapies.

Summary: Harvard Medical School researchers have developed an AI tool called TxGNN that can identify existing drugs as potential treatments for over 17,000 diseases, including many rare conditions currently lacking therapies.

Estimated reading time: 7 minutes

Addressing the Rare Disease Treatment Gap

Rare diseases collectively affect about 300 million people globally, yet only 5 to 7 percent of these conditions have an FDA-approved drug. This leaves a vast majority of rare disease patients with limited or no treatment options.

Marinka Zitnik, assistant professor of biomedical informatics in the Blavatnik Institute at HMS and lead researcher on the project, explains the significance of their work: “With this tool we aim to identify new therapies across the disease spectrum but when it comes to rare, ultrarare, and neglected conditions, we foresee this model could help close, or at least narrow, a gap that creates serious health disparities.”

The TxGNN model represents a significant advancement in AI-driven drug discovery, particularly for rare diseases:

  1. It is the first AI model developed specifically for identifying drug candidates for rare and untreated diseases.
  2. It can handle the largest number of diseases (over 17,000) of any single AI model to date.
  3. The tool is nearly 50 percent better at identifying drug candidates compared to leading AI models for drug repurposing.
  4. It is 35 percent more accurate in predicting drug contraindications.

How TxGNN Works

TxGNN employs a unique approach to drug repurposing:

  1. It identifies shared features across multiple diseases, such as common genomic aberrations.
  2. The model extrapolates from well-understood diseases with known treatments to poorly understood ones without treatments.
  3. It was trained on vast amounts of data, including DNA information, cell signaling, gene activity levels, and clinical notes.

Zitnik highlights the model’s potential: “This is precisely where we see the promise of AI in reducing the global disease burden, in finding new uses for existing drugs, which is also a faster and more cost-effective way to develop therapies than designing new drugs from scratch.”

Advantages of Drug Repurposing

Repurposing existing drugs offers several benefits:

  1. It relies on medicines with well-understood safety profiles.
  2. These drugs have already gone through regulatory approval processes.
  3. It can be faster and more cost-effective than developing new drugs from scratch.

The researchers note that nearly 30 percent of FDA-approved drugs have acquired at least one additional treatment indication following initial approval, with many gaining tens of additional indications over time.

Transparency and Clinical Application

A key feature of TxGNN is its ability to explain the rationale behind its drug recommendations. This transparency can increase physician confidence in the model’s suggestions.

The researchers have made the tool available for free and want to encourage clinician-scientists to use it in their search for new therapies, especially for conditions with no or with limited treatment options.

While the model’s recommendations would require additional evaluation for dosing and delivery timing, its unprecedented capacity could significantly expedite the drug repurposing process.

Questions and Future Directions

As with any new technology, several questions arise:

  1. How will regulatory bodies approach drugs repurposed through AI recommendations?
  2. What steps are needed to validate the model’s suggestions in clinical settings?
  3. How can we ensure equitable access to treatments identified through this method?

The research team is already collaborating with several rare disease foundations to help identify possible treatments, marking the beginning of what could be a new era in rare disease therapy development.

Quiz

  1. What percentage of rare diseases currently have an FDA-approved drug?
  2. How many diseases can the TxGNN model handle?
  3. Compared to leading AI models for drug repurposing, how much better is TxGNN at identifying drug candidates?

Answer Key:

  1. 5 to 7 percent
  2. Over 17,000
  3. Nearly 50 percent better

Further Reading:

  1. TxGNN Explorer: http://txgnn.org/
  2. National Organization for Rare Disorders (NORD): https://rarediseases.org/
  3. FDA’s Orphan Drug Designation program: https://www.fda.gov/industry/developing-products-rare-diseases-conditions
  4. NIH’s National Center for Advancing Translational Sciences: https://ncats.nih.gov/

Glossary of Terms:

  1. Drug Repurposing: The process of identifying new uses for existing approved drugs.
  2. Rare Disease: A condition that affects fewer than 200,000 people in the United States.
  3. TxGNN: The AI model developed by Harvard researchers for drug repurposing.
  4. Off-label Use: The practice of prescribing drugs for conditions other than those for which they were approved.
  5. Contraindication: A condition or factor that serves as a reason to withhold a certain medical treatment due to potential harm.

Enjoy this story? Get our newsletter! https://scienceblog.substack.com


Substack subscription form sign up