VDLIN: A Deep Learning-Based Platform for Methylcobalamin-Inspired Immunomodulatory Compound Screening

During the COVID-19 pandemic, methylcobalamin (MCB), an active form of Vitamin B12 (VB12), showed therapeutic potential in mitigating the cytokine storm associated with SARS-CoV-2 Infection. While MCB's significant anti-inflammatory properties are confirmed, it is also observed that its treatment may impair macrophage-mediated innate immune responses. Comprehensive RNA-seq, ATAC-seq, and CUT&Tag analyses revealed that MCB reduces inflammation and weakens innate immunity by limiting chromatin accessibility at NF-κB and EGR1 binding sites, leading to decreased IFNB1 production and enhanced viral immune evasion. To address this challenge, a deep learning model, VDLIN (Vitamin B12-derived Deep Learning for Innate Immunity), is developed to identify compounds capable of both suppressing inflammation and boosting innate immunity. As anticipated, VDLIN identified a novel compound, "Co7," which retains MCB's strong anti-inflammatory effects while also enhancing immune activation via the TLR4 signaling pathway. Co7 thus emerges as a promising therapeutic candidate, offering advantages over MCB by balancing anti-inflammatory and immune-stimulatory functions. Taken together, this study sheds light on the intricate interplay between chromatin dynamics and immune regulation, presenting new opportunities for therapeutic interventions in inflammatory diseases and SARS-CoV-2 Infection.

EGR1; NF‐κB; TLR4; Vitamin B12 (VB12); convolutional neural network; inflammatory response; innate immunity.