Personalized Swiss Sepsis Study: Detection and modelling of sepsis using machine learning to analyze continuous ICU monitoring, laboratory, microbiology, and -omics data for personalized sepsis management (PSSS). – PHRT
Personalized Swiss Sepsis Study: Detection and modelling of sepsis using machine learning to analyze continuous ICU monitoring, laboratory, microbiology, and -omics data for personalized sepsis management (PSSS).
Sepsis is associated with high morbidity and mortality. Early diagnosis and treatment can significantly influence the heterogeneous outcome. Within this collaborative network of all Swiss University Hospitals, Universities and ETH Zürich, we will generate a highly interoperable research network on novel digital and molecular –omics based biomarkers with the final goal to recognize a bacterial sepsis earlier and to predict its course more precisely than currently possible for an individual patient.
The PSSS Driver project aims to build an interoperable infrastructure
among the intensive care units of the Swiss university hospitals and several research groups, to gather complex information on the host and pathogen during the entire course of a sepsis. The integration of continuous monitoring data from intensive care units will result in digital biomarkers. Combined with the molecular data from bacterial pathogens (metagenomics and whole genome sequencing) and from the host (metabolomics, immunophenotyping, genotyping), new avenues for sepsis research will emerge. These very comprehensive and complex data will be combined via the SPHN data hubs to enable multi-dimensional analyses through machine learning. The goal is to recognize a bacterial sepsis earlier and to predict its course more precisely than currently possible for an individual patient.
The heterogeneous course of sepsis and associated fatal outcome may greatly benefit from a personalized approach in the diagnostics and treatment. Due to low sensitivity and specificity of current biomarkers, sepsis is recognized relatively late, which leads to a reduced efficacy of antibiotic treatment and high mortality. Therefore, novel digital, molecular and hybrid biomarker will help to (i) recognize sepsis at a much earlier state and (ii) predict the likelihood of mortality. Critical for exploring and validating novel types of biomarkers will be sufficiently large study cohorts including multiple centers. The proposal offers the opportunity to (i) form a collaborative network with well-established research groups in the field and (ii) reach sufficient patient numbers within a few years, and (iii) to statistically identify biomarkers that hold the potential to drastically improve the early diagnosis of sepsis and prediction of mortality due to sepsis. The analysis of these complex multidimensional datasets requires specific expertise in bioinformatics, statistics and machine learning – the PHRT will also help to improve the overall data quality and integrate complex -omics data back into routine clinical workflows. To conclude, the Personalized Swiss Sepsis Study will put Switzerland at the forefront of personalized diagnostic and treatment research on sepsis in the world.
Bacterial infection progressing to sepsis is associated with high morbidity, mortality, reduction of quality of life in survivors and health care costs. The course and outcome of sepsis is highly heterogeneous and depends on the causative pathogen and varies from patient to patient. The individual outcome is significantly influenced by various complex host- and pathogen-related factors. Increasing rates of multi-drug resistant bacteria further complicate the diagnostic process and clinical management and may lead to treatment failure. Therefore, patients with sepsis would greatly benefit from personalized diagnostic assessment and treatment strategies evaluating and integrating the host and the pathogen.
Patents / Startups
Weis, C., Cuénod, A., Rieck, B. et al. Direct antimicrobial resistance prediction from clinical MALDI-TOF mass spectra using machine learning. Nat Med 28, 164–174 (2022). https://doi.org/10.1038/s41591-021-01619-9
Michael Moor, Nicolas Bennett, Drago Plečko, Max Horn, Bastian Rieck, Nicolai Meinshausen, Peter Bühlmann, Karsten Borgwardt, Predicting sepsis using deep learning across international sites: a retrospective development and validation study, eClinicalMedicine, Volume 62, 2023, 102124, ISSN 2589-5370, https://doi.org/10.1016/j.eclinm.2023.102124.