Combining Deep Phenotyping of Serum Proteomics and Clinical Data via Machine Learning for COVID-19 Biomarker Discovery

Abstract

the persistence of long-term coronavirus-induced disease 2019 (COVID-19) sequelae demands better insights into its natural history. Therefore, it is crucial to discover the biomarkers of disease outcome to improve clinical practice. In this study, 160 COVID-19 patients were enrolled, of whom 80 had a “non-severe” and 80 had a “severe” outcome. Sera were analyzed by proximity extension assay (PEA) to assess 274 unique proteins associated with inflammation, cardiometabolic, and neurologic diseases. The main clinical and hematochemical data associated with disease outcome were grouped with serological data to form a dataset for the supervised machine learning techniques. We identified nine proteins (i.e., CD200R1, MCP1, MCP3, IL6, LTBP2, MATN3, TRANCE, α2-MRAP, and KIT) that contributed to the correct classification of COVID-19 disease severity when combined with relative neutrophil and lymphocyte counts. By analyzing PEA, clinical and hematochemical data with statistical methods that were able to handle many variables in the presence of a relatively small sample size, we identified nine potential serum biomarkers of a “severe” outcome. Most of these were confirmed by literature data. Importantly, we found three biomarkers associated with central nervous system pathologies and protective factors, which were downregulated in the most severe cases.

Benchmarking of Machine Learning classifiers on plasma proteomic for COVID-19 severity prediction through interpretable artificial intelligence

“…To this end, AI could be used to provide early hints or useful insights regarding the disease progression and the impact of various factors, including the identification of potential causal factors. Albeit a multitude of different AI approaches are being applied in the aforementioned datasets for COVID-19 (e.g., Random Forest, Logistic regression) [9] , there are still several lingering caveats considering predominantly the lack of interpretability and explainability (“black box” challenge) [33] . Acknowledging these hurdles, in the herein work we present a benchmarking pipeline for various ML classifiers based on COVID-19 plasma proteomics (3 datasets based on Olink PEA technology encompassing detailed clinical covariates) engaging an “interpretable” AI approach [34] .…”

“… Dimensionality reduction methods Description Recursive feature extraction (RFE) analysis [9] Eliminates the features that less correlated with the target variable. Feature selection through GINI Index [9] The features are selected based on GINI Index, a measure that calculates the contribution of each feature to the prediction of the output. Recursive Feature Elimination [10] It is a feature importance method that keeps the most significant features for the chosen ML model.…”

“…While this was not the typical case, some works also focus on the respective models' interpretability. Indicatively, SHAP values were used by Beltrami et al [9] , to depict the impact of the training features in the outcome model. Moreover, coefficients are calculated as an explanation method as the mean decrease in the Gini method [12] and minimal-optimal variables method [13] .…”

Extracellular vesicle features are associated with COVID‐19 severity

“…From a clinical point of view, there is a need to identify early biomarkers of adverse outcomes that could help clinicians rapidly discern, among patients with SARS-CoV2 infection and pneumonia, those that will evolve into a critical disease or succumb to COVID-19. For this purpose, we have already employed targeted 7 or high-throughput serum proteomics analyses, 8 demonstrating that inflammatory markers, including interleukin-6, are among the strongest predictors of patient outcome, even when analysed in combination with routine haematological, blood chemistry, demographic and clinical data of enrolled patients. 8 Extracellular vesicles (EV) are cell fragments enclosed in the plasma membrane that are produced in response to a variety of physiological and pathological stimuli.…”

“…For this purpose, we have already employed targeted 7 or high-throughput serum proteomics analyses, 8 demonstrating that inflammatory markers, including interleukin-6, are among the strongest predictors of patient outcome, even when analysed in combination with routine haematological, blood chemistry, demographic and clinical data of enrolled patients. 8 Extracellular vesicles (EV) are cell fragments enclosed in the plasma membrane that are produced in response to a variety of physiological and pathological stimuli. EV can be divided into three broad subgroups (apoptotic bodies, microvesicles and exosomes) based on their size, which partly reflects their mechanism of origin.…”