Artificial Intelligence-based Diagnosis of bloodstream Infection from digitized BOttom-up Proteomes

Sepsis is a syndromic response to bloodstream infection (BSI) that affects 49 million people globally each year and causes 11 million deaths annually. Standard of care (SoC) for diagnosing BSI relies on MALDI-TOF mass spectrometry for pathogen identification (ID) while antibiotic susceptibility testing (AST) is deduced from growth inhibition techniques, at the best within a turnaround time of 24 hrs. There is thus a need for rapid, comprehensive, and affordable diagnosis solution that can provide both pathogen ID and a phenotypic resistance profile. AIDIBOP’s project aims to conceive and validate in a hospital setting a streamlined proteomic sampling assay using the high resolution mass spectrometry DIA mode of acquisition then data processing based on artificial intelligence (AI) for directly inferring i) micro-organism(s) identity, ii) the detection of proteins involved in antimicrobial resistance, iii) the prediction of antimicrobial resistance levels to antibiotics used in first line for BSI stewardship (“pseudo MIC”). The goal is to achieve this with a turnaround time of less than 60 min including the sample preparation, that would make AIDIBOP the most rapid and comprehensive BSI testing solution.