– Pierre Mahé (bioMérieux) : Machine learning and clinical microbiology : some applications for in vitro diagnostics

Carte non disponible

Date/heure
Date(s) - 29 mai 2015

Catégories Pas de Catégories


Title : Machine learning and clinical microbiology : some applications for in vitro diagnostics\n\nAbstract : High-throughput technologies like mass-spectrometry and next-generation sequencing offer powerful means to characterize microorganisms, and are receiving an increasing attention in the field of clinical microbiology. They are in particular increasingly accepted in the context in vitro diagnostics, where machine learning techniques become key to cast the large and/or complex volumes of data produced into a high-level result, ultimately actionable by a clinician to prescribe a therapy.\n \nIn this talk I will present two contributions recently made by bioMérieux in this area. The first topic I will discuss is related to the task of detecting and identifying polymicrobial samples from a MALDI-TOF mass spectrum, for the purpose of which we rely on a penalized non-negative linear regression framework. I will then discuss how large-scale machine learning methods can be useful to estimate the constitution of the bacterial flora present in a sample from next-generation sequencing reads, using techniques inherited from the text processing literature.[