The structural context of mutations in proteins predicts their effect on antibiotic resistance.

Integrating protein structural context enables 96.5% accurate prediction of resistance-conferring mutations in M. tuberculosis proteins.

• Why it matters: Understanding how mutations lead to antibiotic resistance is crucial for developing better diagnostics and treatments, but current sequence-based methods overlook structural signals that may improve predictions.
• What they did: The study curated structural annotations from crystallography and AlphaFold predictions for over 31,000 M. tuberculosis isolates and trained a supervised classifier using 3D mutation distances to known resistance sites.
• The result: The classifier achieved an F1 score of 96.5% in identifying resistance-conferring mutations, demonstrating that protein structure, even when AI-predicted, provides valuable information for resistance prediction.