Secondary Analysis of Clinical Bulk RNA-Sequencing Samples from Dengue Virus-Infected Patients Reveals EGFR, mTOR, and VEGF-A as Potential Therapeutic Targets

1. Introduction

Dengue virus, a member of the Flavivirus genus, consists of four distinct subtypes and is transmitted through a mosquito vector. This arthropod-borne virus causes an estimated 100-400 million clinical infections annually (Roy & Bhattacharjee, 2021), of which only a fraction are reported (Malik et al., 2023). In over 90% of cases, dengue fever is associated with mild symptoms and does not require hospitalization. Although the fatality rate of hemorrhagic severe dengue is relatively low (1-5%), its increasing prevalence remains a global concern (Witte et al., 2024). Climate change allows the mosquito vectors of DENV (primarily Aedes aegypti) to expand into new geographical regions (Hales et al., 2002; Wilder-Smith et al., 2019; Codeco et al., 2022). Further spread is facilitated by urbanization, travel, and globalization (Lee et al., 2021; Harish et al., 2024; Poongavanan et al., 2024).

Treatment of DENV infection generally involves supportive therapy to replace lost plasma volume, with blood transfusions recommended for patients with severe hemorrhaging (Tayal et al., 2023). In parallel, anti-inflammatory therapeutics aid in pain management but can aggravate hemorrhaging (Deen & Seidlein, 2019; Kellstein & Fernandes, 2019). Currently, there are no approved disease-specific drugs or antivirals for treating dengue, although several are being investigated (Dussupt et al., 2020; Martins et al., 2022; Natali et al., 2021).

Previous studies have generated bulk RNA-sequencing data to identify transcriptional markers, mechanisms of disease, and potential host-based therapeutic targets for DENV (Al Noman et al., 2025; Pozzi et al., 2020; Mehta et al., 2024). However, relatively few studies have been conducted on clinical samples collected directly from infected patients (Suppiah et al., 2025; Tsalik et al., 2021). This study aims to analyze high-throughput bulk RNA-sequencing data from peripheral blood mononuclear cells (PBMCs) to identify relevant differentially expressed genes (DEGs), associated cellular pathways, immune cell counts, and existing drug therapies that may be repurposed as potential treatment options for DENV infection.

2. Description

Public RNA-sequencing (RNA-seq) data generated from clinical samples were acquired from the National Center for Biotechnology Information (NCBI) Gene Expression Omnibus (GEO) database (Clough et al., 2024). Fifty-seven studies met the inclusion criteria and were manually reviewed. Samples that did not consist of PBMCs from acutely infected patients were excluded.

Our analysis consisted of two studies (Suppiah et al., 2025; Tsalik et al., 2021) characterizing the host transcriptional response to dengue infection and contained a total of 65 bulk RNA-sequencing samples. We began by performing quality control (FastQC) and trimming (TrimGalore!), mapping and quantification (Salmon), and differential expression (edgeR) between dengue-infected and healthy control patients. The comparison yielded 5921 (2413 upregulated and 3508 downregulated) differentially expressed genes that surpassed our significance threshold of FDR-corrected p-value ≤ 0.05 (Figure 1A).

To assess transcriptional separation between groups, we performed PCA on the top 500 most variable genes across all 65 samples following TMM normalization. While we observed some stratification between batches, PC1 explained 18.2% of the total variance, and there was visible clustering of the control and dengue-infected cohorts, indicating that DENV infection is the dominant source of transcriptional signaling in this dataset (Figure 1B).

We then evaluated pathway-level perturbations using the signaling pathway impact analysis (SPIA) algorithm and the KEGG and Reactome databases. We found forty-three pathways that reached nominal significance (p ≤ 0.05) across both databases, although none maintained significance after multiple hypothesis correction. Given the exploratory nature of this secondary analysis, we report all 10 nominally significant KEGG pathways with strong biological coherence, which should be interpreted as useful for hypothesis generation (Figure 1C). The most significant pathways were related to the cell cycle, the immune system, and metabolism.

We applied CIBERSORTx, a machine learning-based immune cell deconvolution platform for bulk RNA-seq data, to our dataset to determine the immune cell populations most responsible for the changes observed in the DEG and SPIA analyses. Our results were analogous with the current scientific understanding of the immune response to dengue infection, including the increased prevalence of macrophages, dendritic cells, CD4+ T cells, and memory B cells. The observed decrease in CD8+ T cells is likely due to immune exhaustion (Chandele et al., 2016; Estrada-Jiménez et al., 2022; Matos et al., 2015) (Figure 1D).

We then applied the Pathways2Targets algorithm, which uses modulated pathways to predict and prioritize therapeutic targets based on disease association, pathway membership, drug tractability, and other metrics. As dengue fever is characterized by an upregulated immune response, we found that immunosuppressive drugs were frequently among the highest-scoring targets (Figure 1E). Notable targets with literature-backed therapeutic potential include epidermal growth factor receptor (EGFR), mechanistic target of rapamycin kinase (mTOR), and vascular endothelial growth factor A (VEGFA). DENV uses the EGFR signaling cascade during its replication cycle (Duran et al., 2017; Chuang et al., 2020; Srikiatkhachorn et al., 2007), and disruption of this pathway has been shown to have antiviral activity in cell culture studies (Smith et al., 2014; Bhakt et al., 2026). The DENV NS5 protein stimulates mTORC2 signaling to counteract apoptosis (Carter et al., 2022), which justifies the evaluation of mTOR inhibitors as antivirals. However, systematic therapeutic suppression of mTOR is complex, as mTOR activation is essential for both innate and adaptive immune processes (Maiese, 2020). While there is no direct interaction between DENV and VEGFA, this host factor is a primary driver of angiogenesis and increased vasopermeability when upregulated. Inhibition of VEGFA activity may halt and consequently reduce the progression from non-severe to severe dengue and can potentially help treat hemorrhagic symptoms without suppressing immune function (Carter et al., 2022).

3. Conclusions

Our secondary analysis of bulk RNA-seq samples from patients with acute dengue infection revealed broad transcriptional responses compared to healthy controls, providing a transcriptomic framework for understanding the pathology of dengue fever and therapeutic targets that could be repurposed for DENV infection, as supported by emerging scientific literature. For all our predicted therapeutic candidates, the direction of observed perturbation does not directly indicate the appropriate therapeutic strategy. Further functional validation is required to determine whether a given target should be pursued through inhibition, restoration of function, or as a disease state marker. The targets reported should be interpreted as computationally produced hypotheses intended to reinforce ongoing experimental attempts to identify targets that can be repurposed for DENV infection.

4. Methods

Following curation of the RNA-seq samples, the corresponding FASTQ files were downloaded from the Sequence Read Archive (SRA) using sra-tools (version 2.10.8). The FASTQ files, associated metadata file, and configuration file were used as inputs to the Snakemake-based Automated Reproducible Modular workflow for preprocessing and differential analysis of RNA-seq data (ARMOR; version 1.5.10) (Orjuela et al., 2019). The ARMOR workflow includes the following steps: trimming sequencing adapters and poor-quality regions of the reads with TrimGalore! (version 0.6.6) (Singh et al., n.d.), calculating quality control metrics with FastQC (Bonney et al., 2024), performing read pseudomapping and quantifying to the human transcriptome (GRCh38 build 98) with Salmon (version 1.4.0) (Yadav et al., 2023), calculating significant differential gene expression using a negative binomial distribution implemented in edgeR (version 4.0.16) (Robinson et al., 2010), and performing Gene Ontology (GO) enrichment against terms in the Molecular Signatures Database (MSigDB) (Aggarwal et al., n.d.) while adjusting for inter-gene correlation using the Camera algorithm (version 3.50.3) (Moncunill et al., n.d.).

Signaling Pathway Impact Analysis algorithm (SPIA; version 2.58.0) was used to enrich DEGs against intracellular signaling pathways from two databases, KEGG and Reactome (Kanehisa et al., 2025; Milacic et al., 2024; Tarca et al., 2009). SPIA accounts for pathway topology and uses permutation and bootstrapping to calculate the level to which each pathway is significantly perturbed in the comparison. The threshold for significant pathways using the SPIA algorithm was FDR p-values < 0.05.

Digital cytometry was performed using CIBERSORTx (Steen et al., 2020) (Newman et al., 2019). Raw gene counts from Salmon were used as the mixture input, and the LM22 signature matrix was used as a reference panel. B-mode batch correction was applied to account for technical variations between the two source datasets. Cell type fractions were estimated in relative mode using 100 permutations for significance testing. Samples with a CIBERSORTx p-value > 0.05 were excluded from downstream analysis. Three samples (SRR12561101, SRR12561253, and SRR12561274) failed deconvolution and were removed prior to statistical comparisons.

Pathways predicted to be significantly modulated by the SPIA were applied to the Pathway2Targets algorithm (Dobbs Spendlove et al., 2023; Gray et al., 2022). The Pathway2Targets algorithm (version 2.2) takes significantly affected pathways, retrieves the protein members of those pathways, searches the Open Targets drug database (Koscielny et al., 2017) for therapeutics known to target the proteins in each pathway, and prioritizes drug targets based on their relevance to the disease according to more than 20 attributes.