FIGURE 2.

pSUFER identifies flaws in enzyme designs that FuncLib can fix. (a) The xyl8.3 backbone is shown in cartoon with regions that failed to exhibit electron density in a crystallographic analysis (PDB entry: 6FHE) shown in wheat. (Inset) Lys306 is flagged by pSUFER since the Lys is buried in a hydrophobic region without countercharge stabilization. Lys306 is proximal to two loops that failed to exhibit electron density and is in close contact with active‐site position Trp305. (b) For comparison, in the case of the high‐efficiency and accurately designed xyl3.1, pSUFER only flags surface‐exposed polar positions and active‐site positions. In both cases, a position was flagged if computational mutation scanning suggested at least six amino acid identities with ΔΔG < 0 at the position. (c) xyl8.3fix shows an improvement of 330‐fold in activity compared to xyl8.3, k _cat/K _M = 226 and 0.69 M⁻¹ s⁻¹, respectively. Data points and standard deviations are based on at least two repetitions. (d) Improvement in system energy following FuncLib design of positions flagged by pSUFER in 62 models of de novo designed enzymes generated by modular assembly and design. On average, 5–6 mutations are introduced, yielding an average improvement of 13 Rosetta energy units (R.e.u.). pSUFER, protein Strain, Unsatisfactoriness, and Frustration findER.