Artificial Neural Networks for Modeling the Interaction between Cytokines Inducing Lymphopenia in Patients with COVID-19

Abstract

Objective: Cytokines induce tissues damage or inflammation due to infection, contributing to host defense through stimulation of hematopoiesis and acute phase immune reactions. The exaggerated synthesis of cytokines or cytokine storm is directly implicated in the critical cases of patients who have been infected with COVID-19. The cytokine storm may promote apoptosis or necrosis of T cells. Recent medical studies show T cell dramatic elimination and exhaustion in COVID-19 patients requiring Intensive Care Unit (ICU) care. But there is no consensus on whether there is a negative correlation between cytokines concentration and lymphopenia. There is not even agreement on which are all the cytokines involved in this process. How do the involved cytokines interact with each other and how do they affect the number of lymphocytes in patients affected by COVID-19?

Our objective has been to design a computational simulation of these interactions for predicting the lymphopenia in patients with COVID-19.

Methods: Taking the data from a meta-analytical study of medical articles carried out with laboratory findings of samples of patients affected by COVID-19, we have designed an artificial neural network (ANN) based on a cascade-correlation algorithm for modeling the process.

Results: The artificial neural network has predicted with a relative error less than 0.01 the influence of the biochemical cascade originated by cytokines in the lymphopenia of the patients affected by COVID-19.