Researchers develop AI-based methods to identify optimal drug-dose combinations to treat tuberculosis.
The spread of tuberculosis (TB) has diminished in the developed world, but it is still prevalent in the developing parts of the world such as in Asia and Africa. The rise of HIV in the 1980’s also saw an increase in TB infections due to the weakened immune systems of patients with HIV. Currently about 1.6 million people die from TB each year, and 10 million people develop active TB infections, which is also contagious.
Tuberculosis is caused by Mycobacterium tuberculosis bacteria and it generally affects the lungs. Individuals can harbor the TB bacteria but show no symptoms. This is referred to as the latent form of TB and an estimated 2 billion people have latent TB. The Centers for Disease Control and Prevention recommend screening for latent TB in people with increased risk for infection that includes those suffering from HIV/AIDS, those from Asia and Africa, or work with patients with TB. Approximately 10 percent of latent TB cases can develop into active infections. Active infections show signs and symptoms in patients and include persistent cough for more than three weeks, unexplained weight loss, night sweats, chills, and loss of appetite. Active TB is also contagious and the TB bacteria are spread through tiny droplets released in the air when a patient coughs or sneezes.
Current treatments for tuberculosis have limitations
Tuberculosis is treatable. However, there are a number of issues with the current treatment methods and they include:
1) Long course of treatment (over 6 to 8 months): patient compliance becomes low over this long treatment course
2) Drug toxicity: some patients develop severe side effects from the prolonged treatment.
3) Drug resistance: Some patients develop resistance to the drugs requiring a change in treatment, which prolongs the treatment period at times to approximately two years. Drug-resistant TB results in a high fatality rate.
Artificial intelligence is being used to help with tuberculosis treatment
Developing new drugs is one approach to tackle this problem. This approach could be a long and costly method. However, a team of researchers at UCLA have developed a different method to address this issue using existing drugs. They have used the power of artificial intelligence drive data analytic methods to determine the optimal drug and dosage combinations required to effectively treat patients in a shorter period of time. This approach is called the “artificial intelligence-parabolic response surface.”
The basic principle underlying the approach is that some drugs work synergistically, which means that these combinations of drugs are more effective than each of the drugs used on their own. The researchers had also determined in earlier studies that changes in dosages do not cause an abrupt change in effectiveness, which allows for finer modulation of drug dosage. Using these principles, the team led by Dr. Chih-Ming Ho and Dr. Marcus A. Horwitz were able to test 15 of these treatment regimens in cell culture and mouse models. They were able to quickly identify three to four drug-dose combinations amongst the possible billion combinations for the 15 drugs currently available to treat TB. The results of their study were published in the May 2019 issue of PLOS ONE journal. Their results suggest a 75 percent reduction in treatment duration using the drug-dosage combinations that were tested in their study.
The team used generic drugs, excluding drugs that were known to confer drug-resistance to patients. Thus, the drugs that were tested in this ‘universal’ treatment regimen are suitable for all types of TB, including the drug-resistant forms, and they act up to five times faster as compared to current standard treatments. The two most potent regimens identified in this study included the drugs clofazimine, bedaquiline, pyrazinamide, and either amoxicillin/clavulanate or delamanid.
Artificial intelligence identified three treatments that cured tuberculosis
Three of the four regimens from this study showed relapse-free cures after just three weeks of treatment. The fourth regimen achieved the same after five weeks of treatment. In contrast, mice treated with the standard regimen showed the presence of active TB bacteria even after six weeks of treatment and required 16 to 20 weeks of treatment to achieve relapse-free cure.
The two most potent treatment regimen includes all currently approved medications. The dosages tested are comparable to dosages needed to treat humans, suggesting easy transfer of this study into clinical trials. Further testing will be required before the treatment is ready for clinical trials.
The drugs used in this study allow for a ‘universal’ treatment regimen to target both the drug-sensitive strains and the drug-resistant strains of the TB bacteria. However, cell culture and animal experiments were carried out with only the drug-sensitive strains. An important follow-up would be to conduct similar experiments with drug-resistant TB bacteria as well.
New treatment is an advance in the fight against tuberculosis
This study, which follows up earlier research carried out by the same team is an important advance in tackling tuberculosis. As Dr. Marcus Horwitz, senior author of the study notes: “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.”
Written by Bhavana Achary, Ph.D.
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)
