Machine learning used to detect tuberculosis strains' drug resistance

A new machine learning model can instantly and accurately predict a tuberculosis strain's potential resistance to first- and second-line drugs, according to a study published in EBioMedicine on April 29.

Approximately 4 percent of the more than 10 million new cases of TB diagnosed each year are resistant to at least two drugs, and 10 percent of those are "extensively drug-resistant" to multiple medications. Traditional drug resistance tests are often unable to detect all resistance-conferring genetic mutations and can take several weeks, during which time the disease can continue to develop resistance to even more antibiotics.

Researchers from Boston-based Harvard Medical School's department of biomedical informatics developed the new program, which predicts a TB strain's resistance to 10 different antibiotics in a tenth of a second, takes a wider variety of genetic mutations into account and shows greater precision than other models. The program will soon be added to the school's online resource for the analysis of TB genomic data, genTB.

More articles about AI:
AI-powered kidney disease diagnostic tool receives FDA 'Breakthrough Device' designation
Novartis, Pfizer, Centene CEOs on AI's limitations in healthcare
FDA exploring use of AI, blockchain to monitor food safety

Copyright © 2022 Becker's Healthcare. All Rights Reserved. Privacy Policy. Cookie Policy. Linking and Reprinting Policy.


Featured Whitepapers

Featured Webinars