InRegPep: a web server for modeling of parallel beta-sheet fibrillar aggregates of amyloidogenic peptides

Predicting the structures of peptide or protein aggregates that assemble into amyloid protofilaments remains one of the most challenging tasks in computational structural biology. Computational difficulties arise mainly from the large size of the modeled systems and the extensive timescales of aggregation processes. The InRegPep web server is designed for the multiscale structural modeling of peptide aggregates that preferentially adopt in-register parallel β-sheet architectures. The method employs the coarse-grained CABS model, which enables the assembly of a large number of aggregates composed of five interacting peptides. The best models are selected using clustering techniques combined with energy-based and model-symmetry criteria. Top-scoring models are subsequently refined at the all-atom level and further assessed using molecular dynamics simulations. Finally, short protofilament structures are assembled for conformationally stable aggregates. The performance of the InRegPep server was evaluated using a benchmark dataset comprising 15 experimentally determined protofilament structures. For 10 of the 15 systems, the best-performing model among the 10 top-scoring predictions achieved an RMSD below 5 Å. The resulting models provide a valuable foundation for subsequent studies, including investigations of conformational polymorphism, mutational design, and downstream experimental or computational validation.