Abstract
<jats:p>Phosphodiesterase 6B (PDE6B) is a catalytic subunit of the rod photoreceptor cGMP phosphodiesterase complex, and pathogenic variation in PDE6B is associated with inherited retinal dystrophy. Missense variants may alter protein function without producing an obvious loss of the native-like fold, creating a need for structural analyses that extend beyond static sequence-to-structure prediction. Here, we evaluated two disease-associated PDE6B substitutions, L228H and T571M, using AlphaFold-based structural screening and an AMBER molecular dynamics workflow, with UCSF ChimeraX used for structural visualization and residue-level interpretation. T571M produced a larger local change in AlphaFold confidence than L228H (ΔpLDDT, -1.96 versus -0.37), prioritizing residue 571 for dynamic analysis while not itself establishing thermodynamic destabilization. Within the sampled short-timescale trajectories, the wild-type (WT) and T571M constructs retained strong structural overlap. Mean backbone root-mean-square deviation increased modestly from 0.602 Å in WT to 0.669 Å in T571M, whereas mean residue-level root-mean-square fluctuation remained nearly unchanged (0.409 Å and 0.414 Å, respectively). The small global difference contrasted with a clearer local interaction signature: T571M retained 82.9% of starting contacts around residue 571 compared with 90.4% in WT, while forming more replacement contacts (19.25 versus 17.43 contacts per frame). These observations are consistent with local contact rewiring following replacement of a polar, hydroxyl-containing threonine by a larger hydrophobic methionine, rather than generalized unfolding of the modeled construct. The results support a mechanistic model in which T571M redistributes local packing and short-timescale motion while preserving the broader structural architecture. Because the present study is limited by one independent trajectory per condition, 30 ps per condition, and a single 201-residue PDE6B catalytic-domain fragment corresponding to full-length residues 471-671, the findings should be interpreted as a focused mechanistic hypothesis that requires longer replicate simulations and functional validation. Nevertheless, the work illustrates how local contact and dynamic signatures can complement static structural prediction in the assessment of retinal-disease variants.</jats:p>