Home Innovation Artificial Intelligence Tulane Researchers Use AI to D...

Tulane Researchers Use AI to Detect Antibiotic Resistance in Bacteria

Artificial Intelligence

Business Honor
08 April, 2025

Tulane's AI-based model improves detection of antibiotic resistance, enabling faster, more accurate treatments.

Researchers at Tulane University have now created a new artificial intelligence-based technique that more precisely identifies genetic indicators of antibiotic resistance in Staphylococcus aureus and Mycobacterium TB, potentially resulting in quicker and more efficient treatments.

A new Group Association Model (GAM) that employs machine learning to find genetic variants linked to medication resistance is presented in a Tulane study. GAM is more adaptable and capable of identifying previously unidentified genetic alterations because it is not dependent on prior understanding of resistance mechanisms, in contrast to standard techniques that may incorrectly associate unrelated mutations with resistance.

Organizations like the WHO currently use techniques of resistance detection that either overlook rare mutations (as with some DNA-based tests) or take too long (as with culture-based testing). In order to uncover genetic alterations that consistently indicate resistance to particular medications, Tulane's approach compares groups of bacterial strains with varying resistance patterns and analyzes full genome sequences.

In order to find important variants associated with resistance, the researchers used GAM on around 4,000 strains of S. aureus and more than 7,000 strains of Mtb. They discovered that GAM significantly decreased false positives, or incorrectly recognized resistance indicators that may result in ineffective treatment, in addition to matching or surpassing the accuracy of the WHO's resistance database.

Combining machine learning with limited or incomplete data enhanced the capacity to predict resistance. In validation studies using clinical samples from China, the AI-improved model outperformed WHO-based methods in predicting resistance to significant front-line drugs.

This is crucial because early resistance discovery can help doctors customize the right treatment before the infection spreads or gets worse. The model may potentially be used to identify other bacteria or even in agriculture, where crop resistance to antibiotics is an issue, because it can detect resistance without the need for expert-defined rules.