Antibiotic resistance is a major public health threat, and clinicians are increasingly confronted with the critical urgency to rapidly select effective antibiotics for patients with severe bacterial infections. However, conventional methods for diagnosing resistance are slow as they involve bacterial culture. A new generation of sequencers has offered the prospect of obtaining diagnoses within a few hours through analysis of the DNA contained in uncultured clinical samples. The objective of this AEx is to explore, in this context, the computational challenges of resistance diagnostics, using a recently developed technique based on ultra-fast nearest neighbor identification among genomes characterized previously. Challenges include the integration of large and heterogeneous genomic and clinical reference data, the deployment of scalable genomic indexes, as well as the deconvolution of signals of individual bacterial species in real clinical samples.
Inria teams involved