Antonio Laus
PC20 – Antonio Laus
CRS4 (Center for Advanced Studies, Research and Development in Sardinia), Pula, Italy
laus@crs4.it
| A Bayesian-optimized, shape-driven oneOPES framework for resolving membrane-dependent nucleation landscapes of human IAPP toward mechanism-guided design |
| Laus Antonio1, Ferro Elsi1, Atzeni Rossano1, Pieroni Enrico1, Valentini Maria1, Massimo Pisu1 1 CRS4 (Center for Advanced Studies, Research and Development in Sardinia), Pula, Italy |
| Abstract Human islet amyloid polypeptide (hIAPP) aggregation is a key molecular event in type 2 diabetes. Yet early nucleation remains difficult to characterize due to heterogeneous conformational ensembles and metastable intermediates, further modulated by membrane proximity. To address this problem, we developed an enhanced-sampling framework based on oneOPES (On-the-fly Probability Enhanced Sampling) to reconstruct the free-energy landscape of hIAPP nucleation in membrane-near and membrane-far conditions [1]. The method is centered on a custom collective variable (CV) designed to encode peptide shape and conformational identity. This variable is built from structurally interpretable descriptors, including hydrogen-bond patterns and side-chain contacts, selected across reference basins spanning beta-like, alpha-like, and disordered states. The descriptor space is then compressed into a one-dimensional discriminant through an HLDA-based formulation further refined by state-specific gating terms, improving local state resolution while reducing degeneracy along the projected coordinate. To increase robustness and transferability, collective-variable construction is coupled to an AI-assisted Bayesian hyperparameter optimization strategy, in which weights, cutoffs, and state-specific thresholds are tuned through a multi-rung workflow with progressively stricter selection criteria. This enables the systematic identification of a balanced and physically meaningful CV that preserves structural interpretability while maximizing discrimination among nucleation-relevant metastable states. We aim to obtain a reliable free-energy landscape describing how hIAPP explores distinct metastable regions in membrane-near and membrane-far conditions, and to identify the key intermediates and barrier regions controlling the onset of aggregation [2]. These metastable states act as mechanistic gatekeepers for rational modulation of nucleation pathways, since selectively stabilizing off-pathway states or destabilizing nucleation-competent intermediates may reshape the associated free-energy barriers. Overall, this work establishes a computational framework that links advanced CV engineering, AI-assisted Bayesian optimization, and enhanced sampling to the mechanistic dissection of hIAPP nucleation and to the identification of structurally actionable states for anti-aggregation design [3]. |
| References [1] Rizzi, V.; Héritier, M.; Piasentin, N.; Aureli, S.; Gervasio, F.L. The Arch from the Stones: Understanding Protein Folding Energy Landscapes via Bioinspired Collective Variables. J. Phys. Chem. Lett. 2025, 16, 9636–9645. https://doi.org/10.1021/acs.jpclett.5c02079 [2] Elenbaas, B.O.W.; Khemtemourian, L.; Killian, J.A.; Sinnige, T. Membrane-Catalyzed Aggregation of Islet Amyloid Polypeptide Is Dominated by Secondary Nucleation. Biochemistry 2022, 61, 1465–1472. https://doi.org/10.1021/acs.biochem.2c00184 [3] Xu, Y.; Maya-Martinez, R.; Guthertz, N.; Heath, G.R.; Manfield, I.W.; Breeze, A.L.; Sobott, F.; Foster, R.; Radford, S.E. Tuning the Rate of Aggregation of hIAPP into Amyloid Using Small-Molecule Modulators of Assembly. Nat. Commun. 2022, 13, 1040. https://doi.org/10.1038/s41467-022-28660-7 |