BARNO: a batch-aware regulatory network optimization framework reveals a RAN-ENO1-NONO regulatory core in melanoma

High-quality melanoma datasets reported by Livnat et al. in 2018 provide an important opportunity to study treatment-associated regulatory programs in patient-derived tumors, yet batch effects often confound regulatory inference and obscure true biological rewiring. Here we present BARNO (Batch-Aware Regulatory Network Optimization), a GENIE3-based framework that penalizes transcription factors dominated by batch-specific signals while preserving biologically coherent regulatory interactions. Applying BARNO to treated-versus-untreated melanoma datasets reconstructed treatment-associated regulatory architectures and identified a convergent RAN-ENO1-NONO regulatory core. Module reduction and survival analysis further showed that this minimal three-gene signature significantly stratifies patient survival in independent TCGA melanoma cohorts. To evaluate the generalizability of the framework, BARNO was additionally applied to CD4⁺ and CD8⁺ T-cell datasets, where consistent regulatory prioritization patterns were observed, supporting the robustness of the batch-aware weighting scheme across distinct cellular contexts. Together, these results demonstrate that BARNO enables batch-aware regulatory network optimization and facilitates the identification of biologically meaningful regulatory programs in heterogeneous single-cell datasets. BARNO is implemented as an open-source R package and is available on GitHub.