A computer model that helps develop tuberculosis treatments has been created. It has also discovered a new treatment method that can reduce the treatment period by more than 3 months. It is expected to be useful in preventing relapses of multidrug-resistant tuberculosis, which occurs when the first-line treatment fails due to relapses and drug resistance.
A research team led by Professor Radocha Sabic from the University of California, San Francisco (UCSF) announced on the 6th in the international journal "Science Translational Medicine" the development of a computer model that shortens the tuberculosis treatment period and prevents relapses through a multidrug regimen.
Tuberculosis is an infectious disease transmitted by the tuberculosis bacteria. It is spread through respiratory secretions like saliva and nasal mucus, and when it infects a person, it usually causes infections in various organs. Most tuberculosis bacteria infect the lungs, and the term tuberculosis is primarily used to refer to pulmonary tuberculosis.
Tuberculosis can be completely cured with antibiotics that kill the tuberculosis bacteria. There are about 10 types of antibiotics used in tuberculosis treatment, and it is possible to cure the disease by using a multidrug regimen of several antibiotics for 4 to 6 months.
However, the long treatment period has led to an increase in cases of relapse into "multidrug-resistant tuberculosis" as patients arbitrarily discontinue treatment. This is because some tuberculosis bacteria that survive without completing antibiotic treatment can reproduce with resistance. Multidrug-resistant tuberculosis increases the treatment period to 18 to 20 months, significantly raising the burden of tuberculosis management.
Professor Sabic noted, "The biggest goal of tuberculosis treatment is to shorten the treatment period. The longer the treatment period, the lower the treatment adherence rate and the greater the risk of developing multidrug-resistant tuberculosis due to relapses."
The UCSF research team used computer modeling to find treatments that shorten the tuberculosis treatment period. Instead of directly experimenting with the effects of a multidrug regimen using several antibiotics, they analyzed big data to find treatment methods.
The research team collected relapse data from 2,965 cases in 18 previous experiments of tuberculosis treatments. They developed a mathematical modeling to predict the relapse possibilities by gathering treatment progress data collected from experiments using various combinations of antibiotics. They also successfully implemented the ability to predict the treatment effects of new antibiotic combinations using artificial intelligence (AI) technology. The tuberculosis treatment development model created in this way was named "AUROC."
The research team found a multidrug regimen consisting of 2 to 3 types of antibiotics using AUROC and identified the optimal antibiotic combination through experiments with mice. As a result, it was found that using pyrazinamide and diarylquinoline antibiotics together allows for the fastest treatment of tuberculosis. The expected tuberculosis treatment period is about 3 to 4 months, which is a maximum of 3 months faster than before.
The research team stated they would improve the computer model developed this time by combining it with clinical trial data to enhance accuracy while reflecting the differences in efficacy based on antibiotic combinations and dosing.
Professor Sabic remarked, "It is not only about finding a new tuberculosis treatment but also about discovering the optimal combination of existing drugs to shorten the treatment period through an innovative approach. I expect this will improve the efficiency of tuberculosis treatment and shorten the duration of clinical trials for new drug development."
Reference material
Science Translational Medicine (2025), DOI: https://doi.org/10.1126/scitranslmed.adi4000