Fig. 6.

The single instance where an AF2 model contains an uncommon feature, not present in the corresponding crystal structure. Local superposition of T1049 (receptor-binding domain of adhesin MrpH) crystal structure (green) and AF2 model (sky blue) ω-loop showing the uncommon Ser116 st-turn iil in the AF2 model that is absent in the experimental structure. Instead, Ser116 in the crystal structure is buried and forms three favorable electrostatic interactions. Ser116 in the modeled structure is exposed to the solvent. The ω-loop is involved in crystal contacts, and superposing the entire AF2 model on the experimental structure reveals that the predicted loop conformation would clash badly with a symmetry-related molecule.