An explainable model of host genetic interactions linked to COVID-19 severity

Abstract

We employed a multifaceted computational strategy to identify the genetic factors contributing to increased risk of severe COVID-19 infection from a Whole Exome Sequencing (WES) dataset of a cohort of 2000 Italian patients. We coupled a stratified k-fold screening, to rank variants more associated with severity, with training of multiple supervised classifiers, to predict severity on the basis of screened features. Feature importance analysis from tree-based models allowed to identify a handful of 16 variants with highest support which, together with age and gender covariates, were found to be most predictive of COVID-19 severity. When tested on a follow-up cohort, our ensemble of models predicted severity with good accuracy (ACC=81.88%; ROC_AUC=96%; MCC=61.55%). Principal Component Analysis (PCA) and clustering of patients on important variants orthogonally identified two groups of individuals with a higher fraction of severe cases. Our model recapitulated a vast literature of emerging molecular mechanisms and genetic factors linked to COVID-19 response and extends previous landmark Genome Wide Association Studies (GWAS). It revealed a network of interplaying genetic signatures converging on established immune system and inflammatory processes linked to viral infection response, such as JAK-STAT, Cytokine, Interleukin, and C-type lectin receptor signaling. It also identified additional processes cross-talking with immune pathways, such as GPCR signalling, which might offer additional opportunities for therapeutic intervention and patient stratification. Publicly available PheWAS datasets revealed that several variants were significantly associated with phenotypic traits such as “Respiratory or thoracic disease”, confirming their link with COVID-19 severity outcome. Taken together, our analysis suggests that curated genetic information can be effectively integrated along with other patient clinical covariates to forecast COVID-19 disease severity and dissect the underlying host genetic mechanisms for personalized medicine treatments.

Combining viral genomics and clinical data to assess risk factors for severe COVID-19 (mortality, ICU admission, or intubation) amongst hospital patients in a large acute UK NHS hospital Trust

“…Indeed, multiple studies have already been conducted identifying potential susceptibility loci in the human genome that may put patients at increased risk of death or other severe outcomes, including mutations in genes linked to immune response, blood clotting and mucus production [72][73][74][75]. In particular, a recent study using machine learning approaches such as XGBoost identified variants from whole exome sequencing associated with severe COVID-19 [76]. These data identified associations between age, gender, and 16 variants linked to immune system and inflammatory processes able to predict severe outcomes with high accuracy.…”

An Explainable Host Genetic Severity Predictor Model for COVID-19 Patients

“…We developed the HGSP model by combining trained decision tree-based models (Random Forest and XGBoost classifiers) from a 5-Fold CV split of the original problem dataset. See a prior study from Onoja et al, 19 for further details. The HGSP combined these models via an ensemble "VotingClassifier" approach from the "sklearn.ensemble" python library module to aggregate the individual classifiers based on their prediction probabilities (soft margin) of the outcome.…”

“…We consider the following case studies for model validation using the 16 In this study, we utilized the HGSP model we developed from a prior study of training decision tree-based models (Random Forest and XGBoost classifiers) combined from across a 5-fold CV (see further details in Onoja et al, 19 ) The HGSP model voting classifier was developed from high-performance machine learning algorithms that have some interpretability abilities due to their recursive tree-based decision system. We used this approach rather than adopting a complex model such as a deep neural network model to minimize the risk of overfitting and avoid the black box approximation of the problem considering their internal model mechanisms are difficult to interpret.…”

“…We adopt this approach to interpret plausible complex genetic interactions that may alongside patients' clinical covariates, interplay with the host severity of the disease outcome. The HGSP model was developed by combining several trained decision tree-based models (Random Forest and XGBoost classifiers) across a 5-fold cross-validation (CV) splitting computational strategy we employed on the original problem dataset to improve the model stability abilities 20 . We further carried out domain interpretations of the genetic variants linked to the severity of the disease via an enrichment database (Transcription, Pathways, Ontologies, Diseases/Drugs, and Cell Types) 27 .…”

“…To this end, this study aimed to use the HGSP model developed from a previous 2000 cohort study of COVID-9 severity among European patients using their WES and covariates information 20 for out-of-sample prediction (3000 cohort WES and covariates information) and importantly, to provide post-hoc model interpretations and explanations that linked disease traits and COVID-19 severity in patients using a genetic dataset.…”