Panel A shows predicted protein-peptide complex models overlaid with the experimentally determined structure. A boxed region highlights residues where prediction alters the orientation of the peptide helix. Panel B shows the X-ray crystal structure of the complex compared with multiple AlphaFold prediction models. Color gradients indicate prediction confidence and show differences in peptide orientation among models.
Comparison of the experimental CID-miniBD X-ray crystal structure with AlphaFold predictions. (A) Top, AlphaFold 2.3.2 predicted structures of fused CID–BD (light gray), similar to the construct we crystallized, and AlphaFold 2.3.2 multimer prediction of the complex with CID and BD as separate chains (dark gray) superimposed on the X-ray crystal structure of the CID–BD complex with chains colored as in Fig. 1. TTLL5 CID 658–811 rendered transparent; residues 812–820 in TTLL5 CID are highlighted by a red box. Our analysis suggests that the incorrect modeling of these residues by AlphaFold result in an incorrect prediction of the position of the BD helix in the CID–BD complex. (B) X-ray crystal structure of the CID–miniBD complex compared with AlphaFold 2.3.2, AlphaFold 2.3.2 multimer, and AlphaFold 3.0.1 predictions. Dotted red line overlaying predicted structures indicates orientation of miniBD. Alphafold prediction structures colored by confidence score. Both AlphaFold 2.3.2 and AlphaFold 2.3.2 multimer assign medium to high confidence to the incorrectly predicted BD helix, with high confidence for the entire miniBD helix assigned by AlphaFold 2.3.2 multimer.