The COVID-19 pandemic continues to pose a significant global health threat. While research efforts focus on finding a cure, a comprehensive understanding of the disease's nature, particularly its similarities and differences with other viral infections, remains limited. SARS-CoV-2 shares genetic similarities with SARS-CoV and exhibits overlapping symptoms with other respiratory illnesses like influenza. However, the investigation of drugs used to treat other infections, such as HIV, in treating COVID-19, hints at potentially unexplored parallels between the functionalities of different viruses and SARS-CoV-2. Understanding these similarities and differences is crucial for advancing our understanding and developing effective treatments. Network medicine offers powerful tools for analyzing complex interdependencies in biological systems, including those between genes, proteins, biological processes, diseases, and drugs. Protein-protein interactions (PPIs) are fundamental to cellular processes, and analyzing PPI networks can shed light on protein functions and signal transduction pathways. Since PPIs are potential drug targets, analyzing interactomes is essential for drug development. Interactomes are characterized by modular hierarchies crucial for efficient information exchange and system function. PPI network analysis has already provided insights into the interactions between viral and human proteins in SARS-CoV-2 infections, informing our understanding of the virus's structure and function, and suggesting drug repurposing strategies. However, a comprehensive comparison of SARS-CoV-2 with other viruses, beyond those biologically similar, is lacking. This study addresses this gap by using a systematic comparison to reveal hidden similarities and differences that could inform network-based applications complementary to traditional approaches. The study uses statistical physics and computational biology techniques to analyze pan-viral patterns across 93 viruses, including SARS-CoV-2, focusing on the virus-human PPIs as an interdependent system.

Previous research has utilized protein-protein interaction networks to investigate the effects of SARS-CoV-2 viral proteins on human cells, enhancing our understanding of COVID-19 and facilitating applications like drug repurposing. Comparing molecular similarities between SARS-CoV-2 and other viral agents allows leveraging existing knowledge on known viruses to predict the effects of this novel virus. Studies have explored the similarities between SARS-CoV-2 and SARS-CoV, as well as its shared symptoms with influenza. However, research also suggests potential parallels between SARS-CoV-2 and other viruses like HIV, prompting investigation into repurposing drugs originally used for different infection types. Recent studies utilizing network medicine approaches have examined virus-host interactions to identify potential drug targets and repurposing strategies. The comparative analysis of SARS-CoV-2 against other zoonotic coronaviruses has revealed pan-viral disease mechanisms. This study builds upon this work by expanding the comparative analysis to a broader range of viruses and employing multiscale analysis.

The study analyzed virus-human PPIs for 93 viruses, including SARS-CoV-2, using data from the STRING and BIOGRID databases. The human interactome (BIOSTR) was created by standardizing protein names from these databases. For each virus, a virus-host interactome was constructed by combining viral-human interactions (with a confidence score ≥ 0.7) with the BIOSTR network. A different approach was needed for SARS-CoV-2 due to its novelty, using data from a published study. The researchers employed a multiscale analysis to investigate the impact of viral interactions on the human PPI network. They used three different scales: 1. **Microscopic analysis:** This involved two approaches: percolation analysis (assessing the giant connected component size after protein removal, representing viral protein inhibition) and perturbation propagation analysis (quantifying the systemic effects of perturbations starting from targeted proteins). 2. **Mesoscale analysis:** This examined changes in the modular and hierarchical organization of the human interactome caused by viral interactions. The researchers used the Louvain method and Bayesian inference of a hierarchical degree-corrected stochastic block model (DCSBM) to quantify changes in the number of modules, modularity, and hierarchical depth. 3. **Macroscopic analysis:** This employed statistical physics of complex networks to analyze macroscopic properties of virus-human PPI networks. The researchers used a Gibbs-like density matrix to define spectral entropy and the Massieu function, which reflect information dynamics within the networks. They quantified changes (δS and δφ) in von Neumann entropy and Massieu function caused by viral perturbations at different temporal scales (β). Further biological analyses involved clustering the viruses based on their shared human protein targets (first and second-order interactors) and enriched Reactome pathways and Gene Ontology terms using the clusterProfiler R package. Finally, they integrated the findings from these analyses using UMAP dimensionality reduction and HDBSCAN clustering to obtain an integrated view of virus clusters.

Microscopic analyses (percolation and perturbation propagation) did not effectively differentiate between the viruses due to the similarity of interactomes. However, mesoscale analysis, focusing on modularity and hierarchy changes, revealed that SARS-CoV-2 impacted modularity similarly to Influenza A and Bunyavirus. Macroscopic analysis using spectral entropy and the Massieu function showed SARS-CoV-2 clustering with Human Respiratory Syncytial virus at small scales and Measles virus at larger scales. Importantly, SARS-CoV-2's cluster was close to those containing viruses known for systemic effects (Herpesvirus, HIV-1), suggesting SARS-CoV-2 exhibits similar properties. Biological pathway enrichment analysis revealed that SARS-CoV-2's direct targets did not cluster with other coronaviruses but showed similarity with Bunyavirus and Reovirus based on shared pathways. Including second-order interactors broadened the clustering to include viruses affecting diverse systems (skin, liver, immune system, nervous system, etc.), suggesting the systemic impact of SARS-CoV-2. Integrated analysis combining physical and biological approaches consistently showed similarities between SARS-CoV-2 and viruses like Herpesvirus, further supporting the systemic nature of COVID-19. The study notes that the limited availability of SARS-CoV-2 PPI data at the time may have influenced the results.

The findings underscore the systemic nature of SARS-CoV-2 infections, challenging the notion of solely localized effects. The multiscale approach employed allowed for a deeper understanding of the virus's effects across various levels of biological organization. The study's success in identifying SARS-CoV-2's similarities with viruses known for systemic effects, such as Herpesvirus and HIV-1, is noteworthy. These similarities suggest potential therapeutic targets and avenues for drug repurposing. The observed overlap in targeted biological pathways, even when direct protein targets differ, highlights the potential for cross-family functional similarities. This information is valuable for drug repurposing efforts, supplementing purely biological approaches. The study's findings offer potential explanations for the wide range of symptoms and organ involvement associated with COVID-19.

This study demonstrates the power of integrating statistical physics and computational biology to analyze viral-host interactions. The multiscale analysis successfully revealed the systemic nature of SARS-CoV-2 infections, identifying unexpected similarities to other viruses known for their systemic effects. The findings highlight the potential for drug repurposing and provide valuable insights for future research on COVID-19 and other viral diseases. Further research should focus on validating these findings experimentally and exploring the identified similarities for therapeutic development.

The study acknowledges that the limited availability of SARS-CoV-2 PPI data at the time of analysis might have influenced the results. Furthermore, the reliance on computational predictions and database information means that experimental validation of the findings is crucial to confirm the proposed relationships and their implications. The mapping of biological interactions to mathematical models, while providing a consistent framework, involves simplifying assumptions that may not fully capture the complexity of the underlying biological processes.