Evaluating approaches for on-the-fly machine learning interatomic potentials for activated mechanisms sampling with the activation-relaxation technique nouveau.

In the last few years, much effort has gone into developing general machine-learning potentials capable of describing interactions for a wide range of structures and phases. Yet, as attention turns to more complex materials, including alloys and disordered and heterogeneous systems, the challenge of providing reliable descriptions for all possible environments becomes ever more costly. In this work, we evaluate the benefits of using specific vs general potentials for the study of activated mechanisms in solid-state materials. More specifically, we test three machine-learning fitting approaches using the moment-tensor potential to reproduce a reference potential when exploring the energy landscape around a vacancy in Stillinger-Weber silicon crystal and silicon-germanium zincblende structures using the activation-relaxation technique nouveau (ARTn). We find that a targeted on-the-fly approach specific to and integrated into ARTn generates the highest precision on the energetics and geometry of activated barriers while remaining cost-effective. This approach expands the types of problems that can be addressed with high-accuracy ML potential.

Energy Transition Pathways for Deep Decarbonization of the Greater Montreal Region: An Energy Optimization Framework.

More than half of the world's population live in cities, and by 2050, it is expected that this proportion will reach almost 68%. These densely populated cities consume more than 75% of the world's primary energy and are responsible for the emission of around 70% of anthropogenic carbon. Providing sustainable energy for the growing demand in cities requires multifaceted planning approach. In this study, we modeled the energy system of the Greater Montreal region to evaluate the impact of different environmental mitigation policies on the energy system of this region over a long-term period (2020-2050). In doing so, we have used the open-source optimization-based model called the Energy-Technology-Environment Model (ETEM). The ETEM is a long-term bottom-up energy model that provides insight into the best options for cities to procure energy, and satisfies useful demands while reducing carbon dioxide (CO2) emissions. Results show that, under a deep decarbonization scenario, the transportation, commercial, and residential sectors will contribute to emission reduction by 6.9, 1.6, and 1 million ton CO2-eq in 2050, respectively, compared with their 2020 levels. This is mainly achieved by (i) replacing fossil fuel cars with electric-based vehicles in private and public transportation sectors; (ii) replacing fossil fuel furnaces with electric heat pumps to satisfy heating demand in buildings; and (iii) improving the efficiency of buildings by isolating walls and roofs.

A Generalized Attraction-Repulsion Potential and Revisited Fragment Library Improves PEP-FOLD Peptide Structure Prediction.

Fast and accurate structure prediction is essential to the study of peptide function, molecular targets, and interactions and has been the subject of considerable efforts in the past decade. In this work, we present improvements to the popular simplified PEP-FOLD technique for small peptide structure prediction. PEP-FOLD originality is threefold: (i) it uses a predetermined structural alphabet, (ii) it uses a sequential algorithm to reconstruct the tridimensional structures of these peptides in a discrete space using a fragment library, and (iii) it assesses the energy of these structures using a coarse-grained representation in which all of the backbone atoms but the α-hydrogen are present, and the side chain corresponds to a unique bead. In former versions of PEP-FOLD, a van der Waals formulation was used for non-bonded interactions, with each side chain being associated with a fixed radius. Here, we explore the relevance of using instead a generalized formulation in which not only the optimal distance of interaction and the energy at this distance are parameters but also the distance at which the potential is zero. This allows each side chain to be associated with a different radius and potential energy shape, depending on its interaction partner, and in principle to make more effective the coarse-grained representation. In addition, the new PEP-FOLD version is associated with an updated library of fragments. We show that these modifications lead to important improvements for many of the problematic targets identified with the former PEP-FOLD version while maintaining already correct predictions. The improvement is in terms of both model ranking and model accuracy. We also compare the PEP-FOLD enhanced version to state-of-the-art techniques for both peptide and structure predictions: APPTest, RaptorX, and AlphaFold2. We find that the new predictions are superior, in particular with respect to the prediction of small ß-targets, to those of APPTest and RaptorX and bring, with its original approach, additional understanding on folded structures, even when less precise than AlphaFold2. With their strong physical influence, the revised structural library and coarse-grained potential offer, however, the means for a deeper understanding of the nature of folding and open a solid basis for studying flexibility and other dynamical properties not accessible to IA structure prediction approaches.

Molecular Interactions of Tannic Acid with Proteins Associated with SARS-CoV-2 Infectivity.

The overall impact of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) on our society is unprecedented. The identification of small natural ligands that could prevent the entry and/or replication of the coronavirus remains a pertinent approach to fight the coronavirus disease (COVID-19) pandemic. Previously, we showed that the phenolic compounds corilagin and 1,3,6-tri-O-galloyl-ß-D-glucose (TGG) inhibit the interaction between the SARS-CoV-2 spike protein receptor binding domain (RBD) and angiotensin-converting enzyme 2 (ACE2), the SARS-CoV-2 target receptor on the cell membrane of the host organism. Building on these promising results, we now assess the effects of these phenolic ligands on two other crucial targets involved in SARS-CoV-2 cell entry and replication, respectively: transmembrane protease serine 2 (TMPRSS2) and 3-chymotrypsin like protease (3CLpro) inhibitors. Since corilagin, TGG, and tannic acid (TA) share many physicochemical and structural properties, we investigate the binding of TA to these targets. In this work, a combination of experimental methods (biochemical inhibition assays, surface plasmon resonance, and quartz crystal microbalance with dissipation monitoring) confirms the potential role of TA in the prevention of SARS-CoV-2 infectivity through the inhibition of extracellular RBD/ACE2 interactions and TMPRSS2 and 3CLpro activity. Moreover, molecular docking prediction followed by dynamic simulation and molecular mechanics Poisson-Boltzmann surface area (MMPBSA) free energy calculation also shows that TA binds to RBD, TMPRSS2, and 3CLpro with higher affinities than TGG and corilagin. Overall, these results suggest that naturally occurring TA is a promising candidate to prevent and inhibit the infectivity of SARS-CoV-2.

The molecular origin of the electrostatic gating of single-molecule field-effect biosensors investigated by molecular dynamics simulations.

Field-effect biosensors (bioFETs) offer a novel way to measure the kinetics of biomolecular events such as protein function and DNA hybridization at the single-molecule level on a wide range of time scales. These devices generate an electrical current whose fluctuations are correlated to the kinetics of the biomolecule under study. BioFETs are indeed highly sensitive to changes in the electrostatic potential (ESP) generated by the biomolecule. Here, using all-atom solvent explicit molecular dynamics simulations, we further investigate the molecular origin of the variation of this ESP for two prototypical cases of proteins or nucleic acids attached to a carbon nanotube bioFET: the function of the lysozyme protein and the hybridization of a 10-nt DNA sequence, as previously done experimentally. Our results show that the ESP changes significantly on the surface of the carbon nanotube as the state of these two biomolecules changes. More precisely, the ESP distributions calculated for these molecular states explain well the magnitude of the conductance fluctuations measured experimentally. The dependence of the ESP with salt concentration is found to agree with the reduced conductance fluctuations observed experimentally for the lysozyme, but to differ for the case of DNA, suggesting that other mechanisms might be at play in this case. Furthermore, we show that the carbon nanotube does not impact significantly the structural stability of the lysozyme, corroborating that the kinetic rates measured using bioFETs are similar to those measured by other techniques. For DNA, we find that the structural ensemble of the single-stranded DNA is significantly impacted by the presence of the nanotube, which, combined with the ESP analysis, suggests a stronger DNA-device interplay. Overall, our simulations strengthen the comprehension of the inner working of field-effect biosensors used for single-molecule kinetics measurements on proteins and nucleic acids.

The energy landscape governs ductility in disordered materials.

Based on their structure, non-crystalline phases can fail in a brittle or ductile fashion. However, the nature of the link between structure and propensity for ductility in disordered materials has remained elusive. Here, based on molecular dynamics simulations of colloidal gels and silica glasses, we investigate how the degree of structural disorder affects the fracture of disordered materials. As expected, we observe that structural disorder results in an increase in ductility. By applying the activation-relaxation technique (an open-ended saddle point search algorithm), we demonstrate that the propensity for ductility is controlled by the topography of the energy landscape. Interestingly, we observe a power-law relationship between the particle non-affine displacement upon fracture and the average local energy barrier. This reveals that the dynamics of the particles upon fracture is encoded in the static energy landscape, i.e., before any load is applied. This relationship is shown to apply to several classes of non-crystalline materials (oxide and metallic glasses, amorphous solid, and colloidal gels), which suggests that it may be a generic feature of disordered materials.

Corilagin and 1,3,6-Tri-O-galloy-ß-D-glucose: potential inhibitors of SARS-CoV-2 variants.

The COVID-19 disease caused by the virus SARS-CoV-2, first detected in December 2019, is still emerging through virus mutations. Although almost under control in some countries due to effective vaccines that are mitigating the worldwide pandemic, the urgency to develop additional vaccines and therapeutic treatments is imperative. In this work, the natural polyphenols corilagin and 1,3,6-tri-O-galloy-ß-d-glucose (TGG) are investigated to determine the structural basis of inhibitor interactions as potential candidates to inhibit SARS-CoV-2 viral entry into target cells. First, the therapeutic potential of the ligands are assessed on the ACE2/wild-type RBD. We first use molecular docking followed by molecular dynamics, to take into account the conformational flexibility that plays a significant role in ligand binding and that cannot be captured using only docking, and then analyze more precisely the affinity of these ligands using MMPBSA binding free energy. We show that both ligands bind to the ACE2/wild-type RBD interface with good affinities which might prevent the ACE2/RBD association. Second, we confirm the potency of these ligands to block the ACE2/RBD association using a combination of surface plasmon resonance and biochemical inhibition assays. These experiments confirm that TGG and, to a lesser extent, corilagin, inhibit the binding of RBD to ACE2. Both experiments and simulations show that the ligands interact preferentially with RBD, while weak binding is observed with ACE2, hence, avoiding potential physiological side-effects induced by the inhibition of ACE2. In addition to the wild-type RBD, we also study numerically three RBD mutations (E484K, N501Y and E484K/N501Y) found in the main SARS-CoV-2 variants of concerns. We find that corilagin could be as effective for RBD/E484K but less effective for the RBD/N501Y and RBD/E484K-N501Y mutants, while TGG strongly binds at relevant locations to all three mutants, demonstrating the significant interest of these molecules as potential inhibitors for variants of SARS-CoV-2.

Polyphenol-Peptide Interactions in Mitigation of Alzheimer's Disease: Role of Biosurface-Induced Aggregation.

Alzheimer's disease (AD) is the most common age-related neurodegenerative disorder, responsible for nearly two-thirds of all dementia cases. In this review, we report the potential AD treatment strategies focusing on natural polyphenol molecules (green chemistry) and more specifically on the inhibition of polyphenol-induced amyloid aggregation/disaggregation pathways: in bulk and on biosurfaces. We discuss how these pathways can potentially alter the structure at the early stages of AD, hence delaying the aggregation of amyloid-ß (Aß) and tau. We also discuss multidisciplinary approaches, combining experimental and modelling methods, that can better characterize the biochemical and biophysical interactions between proteins and phenolic ligands. In addition to the surface-induced aggregation, which can occur on surfaces where protein can interact with other proteins and polyphenols, we suggest a new concept referred as "confinement stability". Here, on the contrary, the adsorption of Aß and tau on biosurfaces other than Aß- and tau-fibrils, e.g., red blood cells, can lead to confinement stability that minimizes the aggregation of Aß and tau. Overall, these mechanisms may participate directly or indirectly in mitigating neurodegenerative diseases, by preventing protein self-association, slowing down the aggregation processes, and delaying the progression of AD.

Finding Reaction Pathways and Transition States: r-ARTn and d-ARTn as an Efficient and Versatile Alternative to String Approaches.

Finding transition states and diffusion pathways is essential to understand the evolution of materials and chemical reactions. Such characterization is hampered by the heavy computation costs associated with exploring energy landscapes at ab initio accuracy. Here, we revisit the activation-relaxation technique (ARTn) to considerably reduce its costs when used with the density functional theory and propose three adapted versions of the algorithm to efficiently (i) explore the energy landscape of complex materials with the knowledge of a single minimum (ARTn); (ii) identify a transition state when two minima or a guess transition state is given (refining ART or r-ART); and (iii) reconstruct complex pathways between two given states (directed ART or d-ART). We show the application of these three variants on benchmark examples and on various complex defects in silicon. For the latter, the presented improvements to ART lead to much more precise transition states while being 2 to 6 times faster than the commonly used string methods such as the climbing image nudged elastic band method (CI-NEB).

Enthalpy-entropy compensation of atomic diffusion originates from softening of low frequency phonons.

Experimental data accumulated over more than 120 years show not only that diffusion coefficients of impurities ordinarily obey the Arrhenius law in crystalline solids, but also that diffusion pre-exponential factors measured in a same solid increase exponentially with activation energies. This so-called compensation effect has been argued to result from a universal positive linear relationship between entropic contributions and energy barriers to diffusion. However, no physical model of entropy has ever been successfully tested against experimental compensation data. Here, we solve this decades-old problem by demonstrating that atomistically computed harmonic vibrational entropic contributions account for most of compensation effects in silicon and aluminum. We then show that, on average, variations of atomic interactions along diffusion reaction paths simultaneously soften low frequency phonons and stiffen high frequency ones; because relative frequency variations are larger in the lower region of the spectrum, softening generally prevails over stiffening and entropy ubiquitously increases with energy.

Is Hydrogen Diffusion along Grain Boundaries Fast or Slow? Atomistic Origin and Mechanistic Modeling.

We perform comprehensive first-principles calculations and kinetic Monte Carlo simulations to explicitly elucidate the distinct roles of grain boundaries (GBs) in affecting hydrogen (H) diffusion in fcc nickel (Ni). We demonstrate the transition between slow and fast H diffusion along the GB with an abrupt change in H diffusivity. Low-angle GBs are shown to comprise isolated high-barrier regions to trap and inhibit H diffusion, with H diffusivity well prescribed by the classical trapping model, while high-angle GBs are shown to provide interconnected low-barrier channels to facilitate H transport. On the basis of the dislocation description of the GB and the Frank-Bilby model, the slow-fast diffusion transition is identified to result from dislocation core overlapping and is accurately predicted. The present Letter provides key mechanistic insights towards interpreting various experimental studies of H diffusion in metals, new critical knowledge for predictive modeling of H embrittlement, and better understanding of the kinetics of H and other interstitial impurities in microstructures.

Mitigating Alzheimer's Disease with Natural Polyphenols: A Review.

According to Alzheimer's Disease International (ADI), nearly 50 million people worldwide were living with dementia in 2017, and this number is expected to triple by 2050. Despite years of research in this field, the root cause and mechanisms responsible for Alzheimer's disease (AD) have not been fully elucidated yet. Moreover, promising preclinical results have repeatedly failed to translate into patient treatments. Until now, none of the molecules targeting AD has successfully passed the Phase III trial. Although natural molecules have been extensively studied, they normally require high concentrations to be effective; alternately, they are too large to cross the blood-brain barrier (BBB). In this review, we report AD treatment strategies, with a virtually exclusive focus on green chemistry (natural phenolic molecules). These include therapeutic strategies for decreasing amyloid-ß (Aß) production, preventing and/or altering Aß aggregation, and reducing oligomers cytotoxicity such as curcumin, (-)-epigallocatechin-3-gallate (EGCG), morin, resveratrol, tannic acid, and other natural green molecules. We also examine whether consideration should be given to potential candidates used outside of medicine and nutrition, through a discussion of two intermediate-sized green molecules, with very similar molecular structures and key properties, which exhibit potential in mitigating Alzheimer's disease.

Algorithmic developments of the kinetic activation-relaxation technique: Accessing long-time kinetics of larger and more complex systems.

In spite of the considerable computer speed increase of the last decades, long-time atomic simulations remain a challenge and most molecular dynamical simulations are limited to 1 µs at the very best in condensed matter and materials science. There is a need, therefore, for accelerated methods that can bridge the gap between the full dynamical description of molecular dynamics and experimentally relevant time scales. This is the goal of the kinetic Activation-Relaxation Technique (k-ART), an off-lattice kinetic Monte-Carlo method with on-the-fly catalog building capabilities based on the topological tool NAUTY and the open-ended search method Activation-Relaxation Technique (ART nouveau) that has been applied with success to the study of long-time kinetics of complex materials, including grain boundaries, alloys, and amorphous materials. We present a number of recent algorithmic additions, including the use of local force calculation, two-level parallelization, improved topological description, and biased sampling and show how they perform on two applications linked to defect diffusion and relaxation after ion bombardement in Si.

Influence of surface vacancy defects on the carburisation of Fe 110 surface by carbon monoxide.

Adsorption and dissociation of gaseous carbon monoxide (CO) on metal surfaces is one of the most frequently occurring processes of carburisation, known as primary initiator of metal dusting corrosion. Among the various factors that can significantly influence the carburisation process are the intrinsic surface defects such as single surface vacancies occurring at high concentrations due to their low formation energy. Intuitively, adsorption and dissociation barriers of CO are expected to be lowered in the vicinity of a surface vacancy, due to the strong attractive interaction between the vacancy and the C atom. Here the adsorption energies and dissociation pathways of CO on clean and defective Fe 110 surface are explored by means of density functional theory. Interestingly, we find that the O adatom, resulting from the CO dissociation, is unstable in the electron-deficit neighbourhood of the vacancy due to its large electron affinity, and raises the barrier of the carburisation pathway. Still, a full comparative study between the clean surface and the vacancy-defected surface reveals that the complete process of carburisation, starting from adsorption to subsurface diffusion of C, is more favourable in the vicinity of a vacancy defect.

Free-Energy Landscape of the Amino-Terminal Fragment of Huntingtin in Aqueous Solution.

The first exon of Huntingtin-a protein with multiple biological functions whose misfolding is related to Huntington's disease-modulates its localization, aggregation, and function within the cell. It is composed of a 17-amino-acid amphipathic segment (Htt17), an amyloidogenic segment of consecutive glutamines (QN), and a proline-rich segment. Htt17 is of fundamental importance: it serves as a membrane anchor to control the localization of huntingtin, it modulates huntingtin's function through posttranslational modifications, and it controls the self-assembly of the amyloidogenic QN segment into oligomers and fibrils. Experimentally, the conformational ensemble of the Htt17 monomer, as well as the impact of the polyglutamine and proline-rich segments, remains, however, mostly uncharacterized at the atomic level due to its intrinsic flexibility. Here, we unveil the free-energy landscape of Htt17, Htt17Q17, and Htt17Q17P11 using Hamiltonian replica exchange combined with well-tempered metadynamics. We characterize the free-energy landscape of these three fragments in terms of a few selected collective variables. Extensive simulations reveal that the free energy of Htt17 is dominated by a broad ensemble of configurations that agree with solution NMR chemical shifts. Addition of Q17 at its carboxy-terminus reduces the extent of the main basin to more extended configurations of Htt17 with lower helix propensity. Also, the aliphatic carbons of Q17 partially sequester the nonpolar amino acids of Htt17. For its part, addition of Q17P11 shifts the overall landscape to a more extended and helical Htt17 stabilized by interactions with Q17 and P11, which almost exclusively form a PPII-helix, as well as by intramolecular H-bonds and salt bridges. Our characterization of Huntingtin's amino-terminus provides insights into the structural origin of its ability to oligomerize and interact with phospholipid bilayers, processes closely linked to the biological functions of this protein.

Probing Potential Energy Surface Exploration Strategies for Complex Systems.

The efficiency of minimum-energy configuration searching algorithms is closely linked to the energy landscape structure of complex systems, yet these algorithms often include a number of steps of which the effect is not always clear. Decoupling these steps and their impacts can allow us to better understand both their role and the nature of complex energy landscape. Here, we consider a family of minimum-energy algorithms based, directly or indirectly, on the well-known Bell-Evans-Polanyi (BEP) principle. Comparing trajectories generated with BEP-based algorithms to kinetically correct off-lattice kinetic Monte Carlo schemes allow us to confirm that the BEP principle does not hold for complex systems since forward and reverse energy barriers are completely uncorrelated. As would be expected, following the lowest available energy barrier leads to rapid trapping. This is why BEP-based methods require also a direct handling of visited basins or barriers. Comparing the efficiency of these methods with a thermodynamical handling of low-energy barriers, we show that most of the efficiency of the BEP-like methods lie first and foremost in the basin management rather than in the BEP-like step.

Probing the Huntingtin 1-17 membrane anchor on a phospholipid bilayer by using all-atom simulations.

Mislocalization and aggregation of the huntingtin protein are related to Huntington's disease. Its first exon-more specifically the first 17 amino acids (Htt17)-is crucial for the physiological and pathological functions of huntingtin. It regulates huntingtin's activity through posttranslational modifications and serves as an anchor to membrane-containing organelles of the cell. Recently, structure and orientation of the Htt17 membrane anchor were determined using a combined solution and solid-state NMR approach. This prompted us to refine this model by investigating the dynamics and thermodynamics of this membrane anchor on a POPC bilayer using all-atom, explicit solvent molecular dynamics and Hamiltonian replica exchange. Our simulations are combined with various experimental measurements to generate a high-resolution atomistic model for the huntingtin Htt17 membrane anchor on a POPC bilayer. More precisely, we observe that the single α-helix structure is more stable in the phospholipid membrane than the NMR model obtained in the presence of dodecylphosphocholine detergent micelles. The resulting Htt17 monomer has its hydrophobic plane oriented parallel to the bilayer surface. Our results further unveil the key residues interacting with the membrane in terms of hydrogen bonds, salt-bridges, and nonpolar contributions. We also observe that Htt17 equilibrates at a well-defined insertion depth and that it perturbs the physical properties-order parameter, thickness, and area per lipid-of the bilayer in a manner that could favor its dimerization. Overall, our observations reinforce and refine the NMR measurements on the Htt17 membrane anchor segment of huntingtin that is of fundamental importance to its biological functions.

Distinct helix propensities and membrane interactions of human and rat IAPP(1-19) monomers in anionic lipid bilayers.

Islet amyloid polypeptide, IAPP or amylin, is a 37-residue peptide hormone coexpressed with insulin by pancreatic ß-cells. The aggregation of human IAPP (hIAPP) into amyloid deposits is associated with type II diabetes. Substantial evidence suggests that the interaction of anionic membranes with hIAPP may facilitate peptide aggregation and the N-terminal 1-19 fragment (IAPP(1-19)) plays an important role in peptide-membrane interaction. As a first step to understand how structural differences between human and rat IAPP peptides in membranes may influence the later oligomerization process, we have investigated the structures and orientations of hIAPP(1-19) and the less toxic rIAPP(1-19) (i.e., the H18R mutant of hIAPP(1-19)) monomers in anionic POPG bilayers by performing replica exchange molecular dynamics (REMD) simulations. On the basis of ∼20 µs REMD simulations started from a random coil conformation of the peptide placed in water, we find that unfolded h(r)IAPP(1-19) can insert into the bilayers and the membrane-bound peptide stays mainly at the lipid head-tail interface. hIAPP(1-19) displays higher helix propensity than rIAPP(1-19), especially in the L12-L16 region. The helix is oriented parallel to the bilayer surface and buried in the membrane 0.3-0.8 nm below the phosphorus atoms, consistent with previous electron paramagnetic resonance data. The helical conformation is an amphiphilic helix with its hydrophilic and hydrophobic faces pointing, respectively, to the lipid head and tail regions. The H18R substitution enhances the electrostatic interactions of IAPP(1-19) with the membrane, while it weakens the intrapeptide interactions crucial for helix formation, thus leading to lower helix propensity of rIAPP(1-19). Implications of our simulation results on the membrane-mediated IAPP(1-19) oligomerization are discussed.

Early oligomerization stages for the non-amyloid component of α-synuclein amyloid.

In recent years, much effort has focused on the early stages of aggregation and the formation of amyloid oligomers. Aggregation processes for these proteins are complex and their non-equilibrium nature makes any experimental study very difficult. Under these conditions, simulations provide a useful alternative for understanding the dynamics of the early stages of oligomerization. Here, we focus on the non-Aß amyloid component (NAC) of the monomer, dimer, and trimer of α-synuclein, an important 35-residue sequence involved in the aggregation and fibrillation of this protein associated with Parkinson's disease. Using Hamiltonian and temperature replica exchange molecular dynamics simulations combined with the coarse grained Optimized Potential for Efficient peptide structure Prediction potential, we identify the role of the various regions and the secondary structures for the onset of oligomerization. For this sequence, we clearly observe the passage from α-helix to ß-sheet, a characteristic transition of amyloid proteins. More precisely, we find that the NAC monomer is highly structured with two α-helical regions, between residues 2-13 and 19-25. As the dimer and trimer form, ß-sheet structures between residues 2-14 and 26-34 appear and rapidly structure the system. The resulting conformations are much more structured than similar dimers and trimers of ß-amyloid and amylin proteins and yet display a strong polymorphism at these early stages of aggregation. In addition to its inherent experimental interest, comparison with other sequences shows that NAC could be a very useful numerical model for understanding the onset of aggregation.