Identifying novel candidate compounds for therapeutic strategies in retinopathy of prematurity via computational drug-gene association analysis

“

TL;DR

This bioinformatics study identified 47 genes associated with retinopathy of prematurity (ROP) and used drug-gene enrichment analysis to predict 50 candidate compounds, with ascorbic acid, simvastatin, and acetylcysteine being the most significant. The top drug classes include antioxidants, anti-diabetics, NSAIDs, and cardiovascular agents, many of which have known safety profiles and could be repurposed for ROP. These findings provide an unbiased, hypothesis-generating framework for future preclinical and clinical studies to validate novel ROP therapies.

Full summary

4,334 CHARS

**Background:** Retinopathy of prematurity (ROP) is the leading cause of preventable childhood blindness worldwide, affecting 31.3% of legally blind children in the US IRIS Registry. Current treatments (laser photocoagulation, anti-VEGF injections) have limitations including myopia, systemic absorption concerns, and reactivation rates. There is a need for novel preventative and therapeutic strategies with favorable safety profiles. This study uses a network medicine approach to identify existing drugs and compounds that could be repurposed for ROP by analyzing drug-gene associations.

**Methods:** The authors queried the NCBI Gene database (as of Dec 2022) for genes associated with ROP from animal models, human studies, and genomic studies. Only studies demonstrating significant associations were included. The resulting gene list was input into the ToppFun function of the ToppGene Suite, which aggregates 22,832 genes and 77,146 drug annotations from five databases (CTD, Drug Bank, Stitch, Broad Institute CMAP up/down). Gene set enrichment analysis (GSEA) identified compounds with significant associations to the ROP gene list using a hypergeometric distribution with Bonferroni adjustment (FDR-adjusted P-value < 0.05). Compounds with significant toxicities or no clinical indications were manually filtered out. Drug-gene and gene-pathway networks were visualized using Cytoscape with prefuse-directed layouts based on edge betweenness and closeness centrality.

**Key Results:** The NCBI query identified 51 unique genetic and proteomic markers associated with ROP, of which 47 had pharmacologic annotations in ToppGene. Nine genes (Casp8, Atf4, Lgals1, Sema3A, Mir223, Malat1, TEAD4, Arg2, Sucnr1) were from animal OIR models; 38 were from human studies, including 10 from human vitreous/retinal tissue (VEGFA, NOS3, IGF1, TNF, IL6, ACE, EPO, ANGPT1, C3, C5). Enrichment analysis identified 6,603 chemical compounds with FDR-adjusted P < 0.05. After filtering, the top 50 most significant compounds were ranked. The top 10 most significant compounds were: ascorbic acid (P = 7.76 × 10⁻²¹, 21 gene hits), simvastatin (P = 1.44 × 10⁻¹⁸, 19 hits), acetylcysteine (P = 1.60 × 10⁻¹⁷, 20 hits), niacin (P = 4.24 × 10⁻¹⁵, 11 hits), ricinelaidic acid (P = 4.48 × 10⁻¹⁵, 10 hits), penicillamine (P = 2.04 × 10⁻¹⁴, 10 hits), curcumin (P = 2.74 × 10⁻¹⁴, 18 hits), losartan (P = 2.88 × 10⁻¹⁴, 11 hits), capsaicin (P = 3.00 × 10⁻¹⁴, 15 hits), and metformin (P = 4.30 × 10⁻¹⁴, 14 hits). The most common drug classes were antioxidants (ascorbic acid, curcumin, N-acetylcysteine, penicillamine, niacin, alpha-lipoic acid), anti-diabetics (metformin, troglitazone, rosiglitazone, pioglitazone), NSAIDs (ibuprofen, diclofenac, celecoxib, aspirin), and cardiovascular drugs (ACE inhibitors like enalapril, ARBs like losartan and telmisartan, calcium channel blockers like nifedipine). Hub genes with high centrality in the drug-gene network included TNF, VEGFA, IL6, NOS3, IGF1, CASP8, HIF1A, KDR, and FLT1. Top gene ontology terms were vasculature development (P = 1.46 × 10⁻²⁵), response to hypoxia (P = 9.97 × 10⁻²³), and angiogenesis (P = 4.62 × 10⁻²¹). Key pathways included HIF-1 signaling, PI3K-Akt signaling, and angiogenesis.

**Clinical Implications:** This computational analysis provides an unbiased, hypothesis-generating list of candidate compounds for ROP that could be repurposed. Antioxidants like curcumin and N-acetylcysteine have known safety in infants and anti-angiogenic effects in preclinical models. Metformin and thiazolidinediones have anti-angiogenic and anti-inflammatory properties. ACE inhibitors and ARBs may modulate the ocular renin-angiotensin system and reduce VEGF/HIF-1α expression. NSAIDs like ketorolac have preliminary clinical evidence for reducing severe ROP. However, the authors emphasize that these predictions require validation in preclinical and clinical studies. Limitations include reliance on existing biomarker data (which may not fully translate from animal models), potential over-representation of commonly prescribed drugs due to more known drug-gene interactions, and the equal weighting of protective vs. risk gene variants. Despite these constraints, this network medicine approach offers a systematic method for drug discovery in complex, multifactorial diseases like ROP.