Combined clinical and genomic data better predicts COVID-19 severity

By Susha Cheriyedath, M.Sc.Mar 28 2021

The current COVID-19 outbreak is one of the worst pandemics in history and has claimed millions of lives across the globe, overwhelmed healthcare systems, and created unprecedented socio-economic ramifications.

The novel severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) emerged in Wuhan, China, in late 2019 and rapidly spread across the world via human-to-human transmission. Infected individuals are either asymptomatic or exhibit a wide range of symptoms ranging from a mild flu-like illness to rapid progression to pneumonia and sometimes death due to respiratory failure. While patients with mild symptoms recover quickly, those with severe COVID-19 require medical intervention, hospitalization, and even ventilation in ICUs.

This heterogeneity in COVID-19 pathogenesis poses challenges for developing treatment protocols. Studies show that severe symptoms of infectious diseases such as COVID-19 are usually associated with host genome variations. Many recent studies, such as genome-wide association studies (GWAS) of severe COVID-19, have revealed the biological mechanisms behind severe COVID-19 requiring hospitalization. In addition to genomic factors, demographics, lifestyles, socio-economic factors, and comorbidities may also affect the exposure, transmission, severity of disease, and viral load. Rigorous analysis of clinical and genomic datasets is essential to find the biomarkers associated with severe COVID-19.

Components of RubricOE, as ensemble ML pipeline extracting stable features from multi-omics data.

*Important notice: medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.

Analyzing clinical as well as genomics data from a large EHR dataset to find the common risk factors in COVID-19 severity

Most studies so far aimed at finding the common risk factors for severe COVID-19 focused either on the clinical or on the genomics factors, but not both of them. To fill this gap, a group of researchers from the US recently analyzed diverse factors from clinical as well as genomics data from a large EHR dataset to find the common risk factors in COVID-19 severity. The study is published in the preprint server, medRxiv*.

The researchers applied AI methods to better interpret the factors that drive the severity of COVID-19. They used the UK BioBank dataset and analyzed both clinical and genomic data of COVID-19 patients, and studied a large prospective cohort of UK patients in the UKBB10 repository. The dataset had a variety of information of about half a million patients related to diagnosis, demographics, lab tests, medications, and genomic information of approximately half a million patients. The data used was from March 16, 2020, to January 21, 2021.

They leveraged positive-unlabeled machine learning algorithms and used a state-of-the-art genomic analysis framework called RubricOE for genomic feature extraction. The results proposed severity prediction algorithms with a high F1 score and provided insights on clinical and genomic factors affecting the severity of the disease.

Although other studies found factors such as smoking, sex (male), age (older than 65), and pre-existing hypertension, cardiovascular disease, diabetes, respiratory diseases, obesity, kidney diseases, immunosuppression, and cancer to be risk factors for severe disease, this study did not find smoking and presence of respiratory diseases to be significant risk factors for the increase in complications due to COVID-19.

Risk factors associated with COVID-19-related mortality is crucial in future research

The study aimed at identifying significant clinical and genomic factors associated with severe response to the SARS-CoV-2 virus and used the data from a large-scale UK EHR dataset. The team used a particular type of machine learning - positive-unlabeled learning - to address the challenge of noisy class label of severity outcome of COVID-19.

According to the authors, the holistic clinico-genomic modeling used in this study can potentially reveal the effect of such factors in a robust way that can be used as biomarkers for better clinical decision making. They plan to investigate further the effect of vaccination and treatment protocols on the severity of COVID-19 by leveraging the inpatient data in the future. Investigating the risk factors associated with COVID-19-related mortality will also be a crucial topic for future research. The authors also reported on how these risk factors have evolved during the last year of the pandemic with respect to significant events such as the emergence of the new variants, including the B.1.1.7 of the SARS-CoV2 virus.

“In the future, we plan to further investigate the effect of treatment protocols and vaccination on COVID-19 severity leveraging the inpatient data.”

*Important notice: medRxiv publishes preliminary scientific reports that are not peer-reviewed and, therefore, should not be regarded as conclusive, guide clinical practice/health-related behavior, or treated as established information.