The Luxembourg Institute of Health (LIH), in collaboration with 15 European and Canadian institutions, has developed a machine learning model to predict hospital mortality after SARS-CoV-2 infection. Published in Nature Communications, this breakthrough could revolutionise personalised medicine and ease the burden on healthcare systems.
In March 2020, the COVIRNA project was launched to create an RNA-based diagnostic test, using artificial intelligence to predict the clinical course of COVID-19. Dr Yvan Devaux explained that the team focused on non-coding RNAs, identified as potential biomarkers. Previously, the consortium had discovered 2,906 long non-coding RNAs (lncRNAs) related to cardiovascular disease, and applied them to COVID-19, targeting inflammation, a typical response to SARS-CoV-2 infection.
The team analysed blood samples from 1,286 COVID-19 patients in Luxembourg, Germany, the United Kingdom and Canada. Three cohorts, totalling 804 patients, served as a basis for selecting the most successful machine learning models, while a fourth cohort of 482 patients validated these models.
The model identified age and lncRNA LEF1-AS1 as the main predictors of hospital mortality. High levels of LEF1-AS1 correlate with reduced risk of mortality. LEF1-AS1 is involved in the proliferation of B and T immune cells and the regulation of inflammation. In addition, LEF1 could play a protective role in COVID-19-related alveolar lesions.
This model promises to improve patient management by differentiating those at high risk of death from those with a high chance of survival. Dr Devaux also mentioned that tests to predict long COVID are underway, especially within the Luxembourg COVALUX cohort.
The COVIRNA consortium, led by LIH and funded by the European Union’s Horizon 2020 programme, brings together 15 partners from 12 European countries. LIH is a public biomedical research organisation specialising in precision health and aims to become a reference in Europe for the application of scientific advances for the benefit of patients.