Researchers Develop AI-Based Methods to Identify Optimal Drug-Dose Combinations for Tuberculosis Treatment
The Global Impact of Tuberculosis
The prevalence of tuberculosis (TB) has decreased in developed nations, yet it remains a significant health challenge in developing regions, particularly in Asia and Africa. The emergence of HIV in the 1980s correlated with an increase in TB infections due to compromised immune systems among HIV-positive individuals. Currently, approximately 1.6 million people succumb to TB each year, with 10 million new active infections reported annually.
Understanding Tuberculosis
Tuberculosis is caused by the Mycobacterium tuberculosis bacteria and primarily affects the lungs. Many individuals can carry the bacteria without exhibiting symptoms, a condition known as latent TB, which affects an estimated 2 billion people worldwide. The Centers for Disease Control and Prevention (CDC) advises screening for latent TB in high-risk populations, including those with HIV/AIDS, individuals from TB-endemic regions, and healthcare workers dealing with TB patients. About 10 percent of latent TB cases may progress to active infections, which present symptoms such as a persistent cough lasting more than three weeks, unexplained weight loss, night sweats, chills, and loss of appetite. Active TB is contagious, spreading through droplets released when an infected person coughs or sneezes.
Limitations of Current Tuberculosis Treatments
While tuberculosis is treatable, existing treatment methods face several challenges:
1. **Long Treatment Duration**: The standard treatment lasts over 6 to 8 months, leading to low patient compliance.
2. **Drug Toxicity**: Some patients experience severe side effects from prolonged treatment.
3. **Drug Resistance**: A subset of patients develops resistance to drugs, necessitating changes in therapy, which can extend treatment duration to nearly two years. Drug-resistant TB is associated with a high fatality rate.
Artificial Intelligence in Tuberculosis Treatment
To address these issues, researchers at UCLA are exploring the use of artificial intelligence (AI) to optimize drug combinations for TB treatment. Instead of developing new drugs, this innovative approach utilizes existing medications. The researchers implemented an AI-driven data analytic method known as the “artificial intelligence-parabolic response surface” to identify effective drug and dosage combinations that can shorten treatment time.
This methodology is based on the principle that certain drugs can work synergistically, making their combined effect greater than the sum of their individual effects. Previous studies indicated that adjustments in drug dosages do not lead to abrupt changes in effectiveness, facilitating precise dosage modulation. The research team, led by Dr. Chih-Ming Ho and Dr. Marcus A. Horwitz, tested 15 treatment regimens in cell cultures and mouse models. They successfully identified three to four optimal drug-dose combinations from the possible billion combinations available for the 15 drugs used to treat TB. Their findings, published in the May 2019 issue of PLOS ONE, suggest a 75 percent reduction in treatment duration when using these identified combinations.
Identified Drug Regimens and Their Efficacy
The team utilized generic drugs while excluding those known to cause drug resistance. The tested regimens are suitable for all TB types, including drug-resistant strains, and demonstrated up to five times faster action compared to standard treatments. The two most effective regimens identified included clofazimine, bedaquiline, pyrazinamide, and either amoxicillin/clavulanate or delamanid.
Remarkably, three of the four regimens led to relapse-free cures after just three weeks of treatment, with the fourth achieving similar results after five weeks. In contrast, mice receiving the standard treatment still showed active TB bacteria after six weeks, requiring 16 to 20 weeks for a relapse-free cure.
Future Directions and Importance of the Study
The two most effective treatment regimens comprise all currently approved medications, and the dosages tested are comparable to those applied in human treatment, suggesting a straightforward transition to clinical trials. However, further testing is necessary before these treatments can advance to human trials.
The current study represents a significant step forward in the fight against tuberculosis, building on previous research by the same team. As Dr. Marcus Horwitz, the study’s senior author, emphasizes, “If our findings are replicated in human studies, patients will be cured much faster, be more likely to adhere to the drug regimen, suffer less toxicity, and be less likely to develop drug-resistant TB.”
References
Clemens DL, Lee BY, Silva A, Dillon BJ, Masleša-Galić S, Nava S, Ding X, Ho CM, Horwitz MA. Artificial intelligence enabled parabolic response surface platform identifies ultra-rapid near-universal TB drug treatment regimens comprising approved drugs. PLoS One. 2019 May 10;14(5)
Tuberculosis fact sheet.