Abstract
The COVID-19 pandemic necessitated rapid therapeutic discovery. This study utilizes transcriptional profiles from SARS-CoV-2 infected human lung epithelial cells (NHBE, A549_ACE2, and Calu3) to identify dysfunctional signaling pathways. By clustering upregulated pathways into "super-GOs" and employing target-based and signature-based drug matching, the researchers identified potential repurposed medications. Key pathways included JAK-STAT and NFkB signaling. Notably, dexamethasone emerged as a top candidate, aligning with subsequent clinical trial successes.
Introduction
Early in the pandemic, the lack of approved therapies led to a focus on drug repurposing. Traditional approaches often relied on phenotypic observations. This research instead analyzes the molecular signaling pathways hijacked by SARS-CoV-2 to identify drugs that can specifically inhibit these cascades. By understanding the "engine room" of the virus's interaction with host cells, researchers can more accurately predict which existing medications might be effective.
Methods
The study integrated several computational techniques to map the host cell response:
- Transcriptional Profiling: Analysis of RNA-seq data (GSE147507) from infected lung cells to see which genes were significantly up- or down-regulated.
- Signaling Network Analysis: Identification of activated pathways using the KEGG database and shortest-path algorithms to find the most impacted signaling routes.
- Super-GO Clustering: Using affinity propagation to group redundant Gene Ontology (GO) terms into 10 distinct "super-GOs," providing a cleaner set of signatures for drug matching.
- Drug Matching: Utilizing DrugBank for target-based matching and the Connectivity Map (CMAP) for signature-based matching to find drugs that reverse the infection's molecular profile.
Results
The analysis highlighted core inflammatory and growth pathways such as MAPK, JAK-STAT, NFkB, and TNF. These pathways are central to the "cytokine storm" observed in severe COVID-19 cases. By looking for commonalities across different cell lines, the researchers filtered out cell-specific noise to find the core molecular response to the virus.
A total of 220 drugs were identified as potential candidates. Dexamethasone was the most frequent hit in the super-GO analysis, appearing 17 times. This finding was particularly significant as it served as a powerful validation of the model; dexamethasone was later clinically proven to significantly reduce mortality in patients requiring respiratory support.
Of the drugs identified in the primary analysis, 31 were already in clinical trials at the time of the study, including tocilizumab and ruxolitinib. Other high-frequency candidates included hydrocortisone and atorvastatin, both of which have been investigated for managing COVID-19 complications.
Discussion and Limitations
While the computational model successfully identified clinically relevant drugs, it relies on in vitro data from lab-grown cells. The physiological response in a living patient (in vivo) may vary significantly. Furthermore, the model identifies drugs based on molecular signatures but does not account for drug toxicity, dosage, or administration routes. For instance, while certain chemotherapy drugs were identified as hits, their immunosuppressive properties might be counterproductive in a viral infection context. Future work aims to integrate patient-specific data to create a more comprehensive "interactome" of the infection.
Summary updated