Immunoprofiles of COVID-19 uniquely differentiated from other viruses: A machine learning investigation of multiplex immunoassay data

Abstract

Abstract Cytokines and chemokines are vital in maintaining a healthy state by efficiently controlling invading microbes. In addition, the dysregulation of these immune mediators can contribute to viral infection pathology. We comprehensively analyzed the profiles of host immunomodulators in response to infections with members of several virus families, particularly if the SARS-CoV-2 infection produces a unique immune profile compared with other viral infections. Multiplex microbead immunoassay results from 219 datasets with a range of viruses were curated systematically. The curated immunoassay data, obtained using Luminex technology, include 35 different viruses in 18 different viral families; this analysis also incorporated data from studies performed in 7 different cell model systems with 28 different sample types. A multivariate statistical analysis was performed with a specific focus involving many investigations (>10), which include the viral families of Coronaviridae, Orthomyxoviridae, Retroviridae, Flaviviridae, and Hantaviridae. A random forest-based classification of the profiles indicates that IL1-RA, C-X-C motif chemokine ligand 9, C-C motif chemokine ligand 4, interferon (IFN)-λ1, IFN-γ-inducing protein 10, and interleukin (IL)-27 are the top immunomodulators among human samples. Similar approaches only between Coronaviridae (COVID-19) and Orthomyxoviridae (influenza A/B) indicated that transforming growth factor-β, IFN-λ1, IL-9, and eotaxin-1 are important features. In particular, the IFN-λ1 protein was implicated as one of the significant immunomodulators differentiating viral family infection. It is evident that Coronaviridae infection, including SARS-CoV-2, induces a unique cytokine–chemokine profile and can lead to specific immunoassays for diagnosing and prognosis of viral diseases based on host immune responses. Alternatively, we can use diagnosing and prognosing. It is also essential to note that meta-analysis-based predictions must be appropriately validated before clinical implementation.