PepSpace: An Automated, Physics-Driven Directed Evolution Platform for the De Novo Design of Specific Peptide Binders and Condensate Modulators

Intrinsically disordered proteins, topologically complex protein surfaces, and biomolecular condensates regulate essential cellular processes but remain difficult to drug with conventional small molecules. Here, we present PepSpace, a physics-driven directed-evolution platform coupling residue-resolution coarse-grained molecular dynamics with a multi-objective genetic algorithm. PepSpace designs peptide binders by rewarding target engagement while penalizing peptide self-association and off-target binding. Evaluated alongside deep-learning generative models across multiple targets, PepSpace-designed peptides achieve predicted interaction strengths up to two orders of magnitude greater than machine-learning designs, while reproducing experimentally observed binding hierarchies and revealing distinct binding modes. For PIEZO1, PepSpace designs 25-mer peptides that preferentially engage a defined 30-residue intracellular beam epitope while suppressing interactions with the remainder 2,547-residues of the protein. For CHERP biomolecular condensates, dual-action peptides weaken native intermolecular contacts and destabilize the condensed phase. PepSpace provides a physically grounded framework for de novo peptide design against dynamic protein targets and biomolecular condensates, with explicit control over specificity, solubility, and collective interactions.