The COVIRNA session at EHMA 2021 highlights the role of predictive models in improving patient outcomes
On Wednesday, 15 September from 13:30-14:45 CEST, the COVIRNA project hosted a session at the EHMA 2021 Annual Conference titled Towards personalised medicine – an innovative diagnostic tool to identify cardiovascular complications in COVID-19 patients. The session examined the benefits of personalised disease management on health systems and discussed how patient-centred care and personalised medicine can improve disease outcomes and enhance patient quality of life. It also looked at some of the most common challenges which arise when sharing data across health institutions and at how AI can be employed to develop predictive models. The session was moderated by Ms Adriana Voicu, the COVIRNA Project Manager.
The session opened with a presentation by Dr Vincent Mooser (Canada Excellence Research Chair in Genomic Medicine – McGill University; Director of the Quebec COVID-19 Biobank) who first elaborated on why personalised care is critical for the treatment of COVID-19 patients. In addition to COVID-19 being the biggest health challenge of the century, there is also wide inter-individual variability in response to virus exposure. Additionally, there is a need to identify individual-specific clinical and molecular signatures to provide optimal care for patients and tailor prevention and treatment approaches.
Dr Mooser then went on to present the BQC19 study, mandated by Fonds de recherche du Québec – Santé and Génome Québec. BQC19 is designed to support research to identify determinants of COVID-19 susceptibility, severity and outcomes, thereby contributing to the public health response to the pandemic. The study, which looks into clinical and molecular signatures of COVID-19, found that the increased level of OAS-1 is associated with reduced risk of susceptibility, hospitalisation or severe symptoms.
Dr Mooser asserted that BQC19 has made progress in getting more data on molecular signatures of COVID-19, including genetic markers, but there is a need for large-scale international collaborations including thousands of properly consented and clinically investigated participants. He concluded his presentation by highlighting the role the COVIRNA project can play in extensive high-quality multi-layer molecular transcriptomics.
Dr Yvan Devaux (Head of the Cardiovascular Research Unit at Luxembourg Institute of Health; COVIRNA Project Coordinator) followed up with a presentation of the COVIRNA project. He started by underlining that COVIRNA Consortium is a large multidisciplinary team, made of experts in different fields such as AI, biomarkers, cardiology, and molecular diagnostics. Dr Devaux then presented the objective of the COVIRNA project, developing a molecular diagnostic kit to predict outcomes of COVID-19 patients, and outlined the approach the project has adopted in achieving the said objective.
Dr Devaux also reflected on the expected impact of the COVIRNA project, speaking about the impact on healthcare providers, patients and the impact the project will have on society. He highlighted the importance of communicating about the project to healthcare providers and the general population to raise awareness of the diagnostic kit, promoting the COVIRNA flyer and the COVIRNA video.
Dr Venkata Satagopam (Senior Research Scientist and Deputy Head of Bioinformatics core at Luxembourg Centre for System Biomedicine, University of Luxembourg) presented some of the challenges in sharing data across health institutions. In his presentation, Dr Satagopam focused in particular on legal and ethical challenges in data sharing, which mostly result from the absence of legal framework and unclear roles and responsibilities of data usage, and technical challenges, which are the result of insufficient knowledge to prepare data for sharing based on the legal framework.
Dr Satagopam then provided an overview of the data-sharing framework adopted in the COVIRNA project, where encrypted data transfers are collected as data and metadata and sent to Data Provenance and Data Information System (DAISY), a GDPR compliance tool. Data is then integrated and analysed and made accessible to the data user. In his closing remarks, Dr Satagopam presented ways to overcome the data sharing challenges, underlying the importance of defining clear data access policy, consulting institutional Data Protection Officer and executing Data Processing and Sharing/Use Agreements as early as possible, as well as enforcing training on GDPR and data stewardship.
Dr Eric Schordan (Chief Business Officer at Firalis) provided a detailed insight into the innovation used by the COVIRNA project. He started by presenting the discovery of FIMICS, a long non-coding RNA panel to better characterise cardiac disorders and toxicities. Dr Schordan then presented the steps in the COVIRNA workflow to identify biomarkers to predict COVID-19 severity, where the first COVIRNA cohorts are tested with FIMICS kit, that is, library preparation followed by capture, sequencing and data analysis, to predict good and bad outcomes for COVID-19 patients, and identify patients with a high risk of suffering from cardiovascular diseases.
Dr Schordan asserted that, while the innovation presented helps in predicting COVID-19 severity, it is not easy to perform, it is not rapid, and it uses Next Generation Sequencing, a tool not easily available. This makes it unsuitable for use in the healthcare setting. To transform the innovation to precision medicine, the lncRNA signature from FIMICS needs to be translated to qPCR assays. qPCR assays will be clinically validated, easy to use, will provide results in less than 24 hours, and will be readily available as all laboratories are equipped with qPCR devices. Dr Schordan concluded his address by saying that the COVIRNA test will be predictive and preventive, ultimately, improving patient quality of life.
Dr Kanita Karaduzovic-Hadziabdic (Assistant Professor at the International University of Sarajevo) presented how AI can be used to build predictive models, reflecting in particular on the work of the COVIRNA project. Dr Karaduzovic-Hadziabdic provided an overview of steps used in developing predictive models using Machine Learning, a sub-discipline in AI where algorithms are trained to learn from data to make predictions, in this case, to diagnose the disease. She highlighted feature selection, which is reducing the number of features and selecting only those useful for the most desired outcome, as one of the most important steps in the workflow. Since COVIRNA data is highly-dimensional transcriptomic data, feature selection is an important part of the process of biomarker discovery.
Following data collection, data is processed by performing data harmonisation and consolidation, i.e. lncRNA data quality checked, normalised, systematic errors removed. This step is followed by the development of a lncRNA signature biomarker, which is done by performing exploratory analysis where univariable ranking is done based on the differential expression to establish which lncRNAs correlate with disease severity and mortality. These predictors are then used to perform feature selection to be used in the building of predictive Machine Learning models. Lastly, biomarker classifiers are built to predict disease severity and risk of mortality in COVID-19 patients.
The session concluded by highlighting the role predictive models play in targeted prevention and treatments of cardiovascular disease, thereby significantly improving patient outcomes. Additionally, the speakers emphasised the fundamental role of data in developing diagnostic tools as well as the importance of international collaboration, which is needed for pooling more information and developing more accurate models.