Deterministic, branch-selective optimization of peptidomimetic scaffolds reveals design principles for targeting TNF-α

Tumor necrosis factor-α (TNF-α) is a central driver of chronic inflammatory diseases, yet the development of non-biologic TNF-α inhibitors remains challenging due to the large, shallow nature of its protein-protein interface. Peptides offer a promising alternative to small molecules; however, rational optimization of complex peptide architectures such as branched or cyclic scaffolds remains a major bottleneck, as conventional approaches often rely on empirical screening or low-resolution mutagenesis. Here, we report a deterministic, branch-selective optimization strategy for a lysine-centered branched peptidomimetic TNF-α inhibitor using On-Demand Array Synthesis and Screening (ODAST). Through three iterative rounds of hypothesis-driven diversification, we systematically resolved the roles of charge identity, charge density, and cooperative aromatic interactions across distinct branches of the scaffold. This process led to the identification of a conserved bidentate anionic recognition motif, paired with aromatic and hydrogen-bonding elements that cooperatively stabilize TNF-α binding. The optimized peptides exhibited >10-fold improvements in apparent affinity and effectively inhibited TNF-α-induced cytotoxicity, with a lead compound displaying low-micromolar cellular IC₅₀ values. Collectively, this work establishes a generalizable framework for rationally optimizing multivalent peptide architectures against challenging protein surfaces and provides design principles applicable beyond TNF-α to Other cytokines and protein-protein interaction targets.