Network and pharmacophore guided and BCL2 and HSP90AA1 targeted drug repurposable approaches against rheumatoid...

Rheumatoid arthritis (RA) raises the incidence of Diffuse large B-cell lymphoma (DLBCL) development, presenting a critical clinical challenge due to the intersection of autoimmunity and malignancy. However, molecular mechanisms underlying RA-mediated DLBCL progression and therapeutic interventions remain poorly elucidated. To address this gap, we developed a novel GCNConv deep learning pipeline to predict key molecular hubs within shared genes of RA and DLBCL networks and validated them through literature mining. Network and pharmacophore-guided virtual screening evaluated 804 off-patent FDA-approved medications for therapeutic repurposing potential against identified hub targets. Our analysis revealed 86 dysregulated genes common to both conditions, with BCL2 and HSP90AA1 emerging as the most significant hubs, demonstrating deep learning scores of 0.766 and 0.748, respectively. The GCNConv model exhibited exceptional predictive performance with R² values of 0.9480, 0.9288, and 0.9117 for training, validation, and test datasets. Pharmacophore-based screening revealed lapatinib and itraconazole as superior BCL2 inhibitors [-7.621 and -7.075] kcal/mol compared to FDA-approved venetoclax [-6.546] kcal/mol, while voriconazole and regadenoson demonstrated robust HSP90AA1 binding [-7.279 and -7.268] kcal/mol, outperforming FDA-approved cromolyn [-5.304] kcal/mol. These findings establish BCL2 and HSP90AA1 as promising dual-purpose therapeutic targets, providing a foundation for addressing both RA management and RA-mediated DLBCL prevention.