European Medical Journal news story
Machine learning models accurately predict TB drug resistance by estimating minimum inhibitory concentrations, offering clinically relevant insights into treatment response and diagnostic precision.
The models were trained on up to 52% of the World Health Organization mutation catalogue data for Mycobacterium tuberculosis. Despite this, they successfully predicted the effects of 97% of graded mutations in the dataset. This suggests that incorporating multiple biological dimensions can enhance model generalisability and accuracy.
The ability to interpret mutation level effects is particularly relevant for understanding resistance mechanisms and guiding therapeutic decisions. The findings also demonstrate that domain informed machine learning approaches can remain interpretable while achieving high diagnostic performance.
Read the full news story here.
For more TB updates, check out the TB CAB Weekly Newsletter (Issue #14, 3 May 2026).
The newsletter is brought to you by the Global TB Community Advisory Board (TB CAB) with the support of Treatment Action Group (TAG) and the European AIDS Treatment Group (EATG). Subscribe to the newsletter here.
Source : European Medical Journal
Get involved
Are you living with HIV/AIDS? Are you part of a community affected by HIV/AIDS and co-infections? Do you work or volunteer in the field? Are you motivated by our cause and interested to support our work?
Subscribe
Stay in the loop and get all the important EATG updates in your inbox with the EATG newsletter. The HIV & co-infections bulletin is your source of handpicked news from the field arriving regularly to your inbox.