Figure 4.

AlphaFold-Multimer de novo model validation and quality assessment of wild-type and variant Myosin-9 dimer models. Structures of the (a) wild-type and (g) variant Gly455S-Ile1626V:Arg1497Gln (also G455S-I1626V:R1497Q) Myosin-9 dimers, colored on the basis of the local QMEANDisCo scores. (b, h) All atom solvation torsion scores are depicted, contributing to QMEAN-based structural validation by evaluating solvent-accessible surface quality, atomic packing, and torsion angle geometry of the wild-type and variant models, respectively. (c, i) Structures of wild-type and the variant Myosin-9 are shown, with each residue colored according to local QMEANDisCo scores in chain A (gray) and chain B (gold), allowing comparison of predicted structural confidence across variants. (d, j) QMEANDisCo score provides a global and local assessment of protein model quality, with values of (d) 0.6 and (j) 0.48 for the homodimer models, respectively (indicated with a red dot). These scores are plotted against the number of residues in the homodimer, reflecting the predicted structure’s overall reliability on the basis of protein size. (e, k) Ramachandran plots illustrate the distribution of backbone dihedral angles (Φ vs Ψ) for individual residues, providing an assessment of sterically allowed conformations and local protein geometry quality. In plots e and k, 92.7% and 88.5% of residues, respectively, fall within favored regions (green), indicating high-quality structural modeling. (f, l) AlphaFold-Multimer de novo models of the (f) wild-type Myosin-9 homodimer and (l) the patient-derived mutant dimer are shown. In the wild-type model, chain A (light teal) and chain B (dark teal) include residues Gly455, Arg1497, and Ile1626. In the mutant model, residues were assigned on the basis of familial segregation: chain A (light brown) contains sequence variants Ser455 and Val1626, whereas chain B (dark brown) includes the Gln1497 variant.