pipette
ENEnglish

Parametric and AI-Assisted Design of Right-Handed Four-Helix Coiled Coils

P. Kluge, A. Useini, F. Shabani, B. Kieslich, A. Kazimir, N. Sträter, C. Lamers, S. Gemming, G. Künze

Preprint

En palabras de los autores

Right-handed, antiparallel -helical coiled coils are rare in nature and remain challenging targets for de novo protein design. Here, we combined parametric backbone generation with deep-learning-based sequence design and structure prediction to systematically explore this underrepresented region of coiled-coil structural space. We compared three ProteinMPNN design strategies: repeat-constrained homotetramer design, unconstrained homotetramer design, and unconstrained heterotetramer design. Removing strict sequence repeat constraints increased the fraction of designs satisfying structural confidence and topology criteria approximately sixfold, from 0.44% to 2.80%, while simultaneous optimization of two distinct peptide sequences yielded a success rate of 3.04%. Selected designs were evaluated for assembly using a split-luciferase complementation assay in mammalian cells, followed by circular dichroism spectroscopy and molecular dynamics simulations. Five of six experimentally characterized designs exhibited predominantly -helical structures, with apparent melting temperatures ranging from 55.2 to 72.5 {degrees}C. The crystal structure of the most stable design, Heteromer-2, determined at 1.92 [A] resolution, confirmed the intended right-handed, antiparallel architecture, with a backbone RMSD of only 0.45 [A] relative to the computational model. Our findings establish a strategy for designing noncanonical coiled-coil architectures by combining precise parametric control of backbone geometry with flexible, AI-assisted sequence optimization.

Resultado principalEl resumen no menciona limitaciones.

Apareció: jueves, 19 de noviembre. bioRxiv. Preprint, todavía sin revisión por pares.

DOI: 10.64898/2026.11.14.766675