File Name : figs1.tiff

Caption : fig. s1. hot-spot engagement of cycdiff-dpo designs for the two ppi targets in the case studies. histograms show the number of designed cyclic peptides versus the number of hotspot residues recovered through peptide–target contacts per design for (a) keap1–nrf2 and (b) spsb2–inos.

File Name : figs2.tiff

Caption : fig. s2. representative lnr complexes incompatible with cyclic peptide scaffolds. (a) pdb 1jrr: the ligand adopts an extended β-sheet conformation that is unlikely to be accommodated by a cyclic peptide of similar length. (b) pdb 4aph: the ligand is buried within the protein core, where cyclization would likely cause steric clashes and poor shape complementarity.

File Name : figs3.tiff

Caption : fig. s3. average intramolecular hydrogen-bond (imhb) count per generated cyclic peptide across five generative models.

File Name : figs4.tiff

Caption : fig. s4. distribution of intramolecular hydrogen-bond (imhb) counts for cyclic peptides in the cycpeptmpdb database, stratified by experimental permeability. peptides with log⁡p_exp > -6 cm/s are classified as high permeability; all others are classified as low permeability. p values were computed using a two-sided mann–whitney u test, with p < 0.05 considered statistically significant. (a) compounds with caco-2 permeability measurements. (b) compounds with pampa permeability measurements.

File Name : figs5.tiff

Caption : fig. s5. boltzmann-weighted imhb counts for cyclic peptides in a membrane-mimicking environment (alpb chloroform). peptides with log p_exp   > −6 cm/s are classified as high permeability; all others are classified as low permeability. p values were computed using a two-sided mann–whitney u test, with p < 0.05 considered statistically significant. (a) compounds with caco-2 permeability measurements. (b) compounds with pampa permeability measurements.