Figure 6. Structural and energetic validation of predicted compensatory complexes.

(a,b) AlphaFold3-derived interfacial contact maps for compensatory 1E6–pMHC complexes. (a) Hot-spot escape mutant Gly4 →Asp rescued by the CDR3α substitution S94 →Y. (b) Flanking escape variant ALGGPDPAAA rescued by the CDR3α substitution D→E. In both cases, AlphaFold3 reproduces the peptide-facing contact patterns predicted by RACER-m, confirming that the compensatory solutions correspond to geometrically feasible TCR–peptide interfaces. (c) Number of low-frustration peaks exceeding a given threshold. The number of local maxima (“peaks”) in the minimally frustrated (low-frustration, green) contact–density profile is plotted as a function of the applied threshold for the wild-type (green), compensatory (blue), and escape (red) complexes. Each peak corresponds to a contiguous cluster of energetically favorable, minimally frustrated interactions along the TCR–peptide interface (see Figure S9–S11 for representative frustration–density profiles and peak identification). At low thresholds, the wild-type structure exhibits the largest number of low-frustration peaks, indicating a broadly distributed network of stabilizing contacts. As the threshold increases, the number of surviving peaks decreases in all cases, with the compensatory complex retaining fewer but more localized low-frustration peaks, consistent with energetic redistribution rather than full restoration of the wild-type stability pattern.