Systems biology and artificial intelligence analysis highlights the pleiotropic effect of IVIg therapy in autoimmune diseases with a predominant role on B cells and complement system

Abstract

Intravenous immunoglobulin (IVIg) is used as treatment for several autoimmune and inflammatory conditions, but its specific mechanisms are not fully understood. Herein, we aimed to evaluate, using systems biology and artificial intelligence techniques, the differences in the pathophysiological pathways of autoimmune and inflammatory conditions that show diverse responses to IVIg treatment. We also intended to determine the targets of IVIg involved in the best treatment response of the evaluated diseases. Our selection and classification of diseases was based on a previously published systematic review, and we performed the disease characterization through manual curation of the literature. Furthermore, we undertook the mechanistic evaluation with artificial neural networks and pathway enrichment analyses. A set of 26 diseases was selected, classified, and compared. Our results indicated that diseases clearly benefiting from IVIg treatment were mainly characterized by deregulated processes in B cells and the complement system. Indeed, our results show that proteins related to B-cell and complement system pathways, which are targeted by IVIg, are involved in the clinical response. In addition, targets related to other immune processes may also play an important role in the IVIg response, supporting its wide range of actions through several mechanisms. Although B-cell responses and complement system have a key role in diseases benefiting from IVIg, protein targets involved in such processes are not necessarily the same in those diseases. Therefore, IVIg appeared to have a pleiotropic effect that may involve the collaborative participation of several proteins. This broad spectrum of targets and 'non-specificity' of IVIg could be key to its efficacy in very different diseases.

Keywords: B cells; IVIg Immunoglobulins; autoimmune diseases; complement system; immunomodulation; inflammatory diseases; intravenous immunoglobulin; mathematical models.

Figure 1

Results from the pathophysiological process grouping independence analysis with respect to different IVIg response clusters. Table (A) and graph (B) show the frequency (%) of conditions in each response cluster associated with each pathophysiological process. * Statistical significant (p< 0.05) result of the chi-squared test indicating dependence between the pathophysiological process group and cluster based in IVIg response.

Figure 2

Protein and functional overlap between the IVIg response-based clusters of diseases. (A) Binary tree based on protein overlap among the disease clusters according to IVIg response. Percentages indicate the arithmetic mean of proportions of protein overlap of each pair of disease clusters or group of clusters in the tree branches; (B) Overlap of protein effectors between diseases with different response to IVIg (Venn diagram created with InteractiVenn (27)) and number of enriched pathways in the three protein sets defined according to the response-based clusters of diseases. the sets were defined in terms of number of proteins and enriched pathways according to the hypergeometric method (FDR q-value< 0.05); (C) Network representation of the 81 pathways enriched in the DB w/o UPB set of proteins (see Table 3 ) considering a Haussdorf distance< 1. IVIg: Intravenous immunoglobulin; only-DB: Protein effectors exclusively present in the IVIg ‘definitely beneficial’ cluster of diseases; DB w/o UPB: Protein effectors present in the IVIg ‘definitely beneficial’ cluster of diseases but not in the IVIg ‘unlikely to provide benefit’ cluster; only-UPB: Protein effectors exclusively present in the IVIg ‘unlikely to provide benefit’ cluster of diseases.

Figure 3

Predicted relationship between each studied immune-related disorder and each pathophysiological process affected by IVIg according to ANN. ANN, Artificial Neural Network; IVIg, Intravenous immunoglobulin.

Figure 4

Predicted relationship between each studied disorder and (A) each complement system protein target of IVIg or (B) each B-cell related IVIg protein target according to ANN. ANN, Artificial Neural Network; IVIg, Intravenous immunoglobulin.