No abstract

Link prediction in artificial intelligence is used to identify missing links or derive future relationships that can occur in complex networks. A link prediction model was developed using the complex heterogeneous biomedical knowledge graph, SemNet, to predict missing links in biomedical literature for drug discovery. A web application visualized knowledge graph embeddings and link prediction results using TransE, CompleX, and RotatE based methods. The link prediction model achieved up to 0.44 hits@10 on the entity prediction tasks. The recent outbreak of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2), also known as COVID-19, served as a case study to demonstrate the efficacy of link prediction modeling for drug discovery. The link prediction algorithm guided identification and ranking of repurposed drug candidates for SARS-CoV-2 primarily by text mining biomedical literature from previous coronaviruses, including SARS and middle east respiratory syndrome (MERS). Repurposed drugs included potential primary SARS-CoV-2 treatment, adjunctive therapies, or therapeutics to treat side effects. The link prediction accuracy for nodes ranked highly for SARS coronavirus was 0.875 as calculated by human in the loop validation on existing COVID-19 specific data sets. Drug classes predicted as highly ranked include anti-inflammatory, nucleoside analogs, protease inhibitors, antimalarials, envelope proteins, and glycoproteins. Examples of highly ranked predicted links to SARS-CoV-2: human leukocyte interferon, recombinant interferon-gamma, cyclosporine, antiviral therapy, zidovudine, chloroquine, vaccination, methotrexate, artemisinin, alkaloids, glycyrrhizic acid, quinine, flavonoids, amprenavir, suramin, complement system proteins, fluoroquinolones, bone marrow transplantation, albuterol, ciprofloxacin, quinolone antibacterial agents, and hydroxymethylglutaryl-CoA reductase inhibitors. Approximately 40% of identified drugs were not previously connected to SARS, such as edetic acid or biotin. In summary, link prediction can effectively suggest repurposed drugs for emergent diseases.

COVID-19 causes significant morbidity and mortality and early intervention is key to minimizing deadly complications. Available treatments, such as monoclonal antibody therapy, may limit complications, but only when given soon after symptom onset. Unfortunately, these treatments are often expensive, in limited supply, require administration within a hospital setting, and should be given before the onset of severe symptoms. These challenges have created the need for early triage of patients likely to develop life-threatening complications. To meet this need, we developed an automated patient risk assessment model using a real-world hospital system dataset with over 17,000 COVIDpositive patients. Specifically, for each COVID-positive patient, we generate a separate risk score for each of four clinical outcomes including death within 30 days, mechanical ventilator use, ICU admission, and any catastrophic event (a superset of dangerous outcomes). We hypothesized that a deep learning binary classification approach can generate these four risk scores from electronic healthcare records data at the time of diagnosis. Our approach achieves significant performance on the four tasks with an area under receiver operating curve (AUROC) for any catastrophic outcome, death within 30 days, ventilator use, and ICU admission of 86.7%, 88.2%, 86.2%, and 87.8%, respectively. In addition, we visualize the sensitivity and specificity of these risk scores to allow clinicians to customize their usage within different clinical outcomes. We believe this work fulfills a clear clinical need for early detection of objective clinical outcomes and can be used for early screening for treatment intervention.

No abstract