Valéry Weber

Semiempirical Molecular Dynamics (SEMD) I: Midpoint-Based Parallel Sparse Matrix-Matrix Multiplication Algorithm for Matrices with Decay.

In this paper, we present a novel, highly efficient, and massively parallel implementation of the sparse matrix-matrix multiplication algorithm inspired by the midpoint method that is suitable for matrices with decay. Compared with the state of the art in sparse matrix-matrix multiplications, the new algorithm heavily exploits data locality, yielding better performance and scalability, approaching a perfect linear scaling up to a process box size equal to a characteristic length that is intrinsic to the matrices. Moreover, the method is able to scale linearly with system size reaching constant time with proportional resources, also regarding memory consumption. We demonstrate how the proposed method can be effectively used for the construction of the density matrix in electronic structure theory, such as Hartree-Fock, density functional theory, and semiempirical Hamiltonians. We present the details of the implementation together with a performance analysis up to 185,193 processes, employing a Hamiltonian matrix generated from a semiempirical NDDO scheme.

Importance of Hydrophilic Hydration and Intramolecular Interactions in the Thermodynamics of Helix–Coil Transition and Helix–Helix Assembly in a Deca-Alanine Peptide

For a model deca-alanine peptide the cavity (ideal hydrophobic) contribution to hydration favors the helix state over extended states and the paired helix bundle in the assembly of two helices. The energetic contributions of attractive protein-solvent interactions are separated into quasi-chemical components consisting of a short-range part arising from interactions with solvent in the first hydration shell and the remaining long-range part that is well described by a Gaussian. In the helix-coil transition, short-range attractive protein-solvent interactions outweigh hydrophobic hydration and favor the extended coil states. Analysis of enthalpic effects shows that it is the favorable hydration of the peptide backbone that favors the unfolded state. Protein intramolecular interactions favor the helix state and are decisive in favoring folding. In the pairing of two helices, the cavity contribution outweighs the short-range attractive protein-water interactions. However, long-range, protein-solvent attractive interactions can either enhance or reverse this trend depending on the mutual orientation of the helices. In helix-helix assembly, change in enthalpy arising from change in attractive protein-solvent interactions favors disassembly. In helix pairing as well, favorable protein intramolecular interactions are found to be as important as hydration effects. Overall, hydrophilic protein-solvent interactions and protein intramolecular interactions are found to play a significant role in the thermodynamics of folding and assembly in the system studied.

Communication: regularizing binding energy distributions and thermodynamics of hydration: theory and application to water modeled with classical and ab initio simulations.

The high-energy tail of the distribution of solute-solvent interaction energies is poorly characterized for condensed systems, but this tail region is of principal interest in determining the excess free energy of the solute. We introduce external fields centered on the solute to modulate the short-range repulsive interaction between the solute and solvent. This regularizes the binding energy distribution and makes it easy to calculate the free energy of the solute with the field. Together with the work done to apply the field in the presence and absence of the solute, we calculate the excess chemical potential of the solute. We present the formal development of this idea and apply it to study liquid water.

Solvation Free Energy of the Peptide Group: Its Model Dependence and Implications for the Additive-Transfer Free-Energy Model of Protein Stability

The group-additive decomposition of the unfolding free energy of a protein in an osmolyte solution relative to that in water poses a fundamental paradox: whereas the decomposition describes the experimental results rather well, theory suggests that a group-additive decomposition of free energies is, in general, not valid. In a step toward resolving this paradox, here we study the peptide-group transfer free energy. We calculate the vacuum-to-solvent (solvation) free energies of (Gly)n and cyclic diglycine (cGG) and analyze the data according to experimental protocol. The solvation free energies of (Gly)n are linear in n, suggesting group additivity. However, the slope interpreted as the free energy of a peptide unit differs from that for cGG scaled by a factor of half, emphasizing the context dependence of solvation. However, the water-to-osmolyte transfer free energies of the peptide unit are relatively independent of the peptide model, as observed experimentally. To understand these observations, a way to assess the contribution to the solvation free energy of solvent-mediated correlation between distinct groups is developed. We show that linearity of solvation free energy with n is a consequence of uniformity of the correlation contributions, with apparent group-additive behavior in the water-to-osmolyte transfer arising due to their cancellation. Implications for inferring molecular mechanisms of solvent effects on protein stability on the basis of the group-additive transfer model are suggested.

Conditional solvation thermodynamics of isoleucine in model peptides and the limitations of the group-transfer model.

The hydration thermodynamics of the amino acid X relative to the reference G (glycine) or the hydration thermodynamics of a small-molecule analog of the side chain of X is often used to model the contribution of X to protein stability and solution thermodynamics. We consider the reasons for successes and limitations of this approach by calculating and comparing the conditional excess free energy, enthalpy, and entropy of hydration of the isoleucine side chain in zwitterionic isoleucine, in extended penta-peptides, and in helical deca-peptides. Butane in gauche conformation serves as a small-molecule analog for the isoleucine side chain. Parsing the hydrophobic and hydrophilic contributions to hydration for the side chain shows that both of these aspects of hydration are context-sensitive. Furthermore, analyzing the solute–solvent interaction contribution to the conditional excess enthalpy of the side chain shows that what is nominally considered a property of the side chain includes entirely nonobvious contributions of the background. The context-sensitivity of hydrophobic and hydrophilic hydration and the conflation of background contributions with energetics attributed to the side chain limit the ability of a single scaling factor, such as the fractional solvent exposure of the group in the protein, to map the component energetic contributions of the model-compound data to their value in the protein. But ignoring the origin of cancellations in the underlying components the group-transfer model may appear to provide a reasonable estimate of the free energy for a given error tolerance.

Early experiences with scientific applications on the IBM Blue Gene/Q supercomputer

We report early experiences with porting highly complex scientific applications to the IBM Blue Gene®/Q platform. In addition, we report our progress in porting performance analysis tools that are deemed to be key in helping users understand massively parallel, massively threaded applications. Porting proved to be quite a smooth process. Although in this early study we did not use the full array of the novel architectural features, we nevertheless obtained quite satisfactory, though preliminary, performance results. Thus, we can safely anticipate impressive further improvements in overall performance once the full capability of the Blue Gene/Q architecture is exploited.

First Experiences with ab initio Molecular Dynamics on OpenPOWER: The Case of CPMD

In this article, we present the algorithmic adaptation and code re-engineering required for porting highly successful and popular planewave codes to next-generation heterogeneous OpenPOWER architectures that foster acceleration and high bandwidth links to GPUs. Here we focus on CPMD as the most representative software for ab initio molecular dynamics simulations. We have ported the construction of the electronic density, the application of the potential to the wavefunctions and the orthogonalization procedure to the GPU. The different GPU kernels consist mainly of fast Fourier transforms (FFT) and basic linear algebra operations (BLAS). The performance of the new implementation obtained on Firestone (POWER8/Tesla) is discussed. We show that the communication between the host and the GPU contributes a large fraction of the total run time. We expect a strong attenuation of the communication bottleneck when the NVLink high-speed interconnect will be available.

Enhanced MPSM3 for applications to quantum biological simulations

Classical molecular dynamics simulations have been the preferred method to cope with the characteristic sizes and time scales of complex life-science systems. However, while classical methods have well known limitations, such as that their accuracy strongly depends on empirical tuning, the practical use of far more accurate methods that rely on quantum Hamiltonians, has been limited by the current efficiency and scalability of sparse matrix-matrix multiplication algorithms used in the self-consistent field equations. In this paper, we show unprecedented massive scalability of a recently presented method, called MPSM3, for sparse matrix-matrix multiplication. The algorithmic basis of the method was presented in a recent publication, while here we describe the algorithmic enhancements that allow us to claim at least one order of magnitude improvement in scalability and time to solution over the state of the art (original MPSM3). We achieve a time to solution for the multiplication of density matrices within the self-consistent field scheme that is approaching the time needed to evaluate energy and forces with classical force-field methods and that is independent from the system size, provided proportional resources. This latest development renders the application of entirely quantum Hamiltonians to systems of several millions of atoms for extended molecular dynamics investigations feasible.

Shedding Light on Lithium/Air Batteries Using Millions of Threads on the BG/Q Supercomputer

In this work, we present a novel parallelization scheme for a highly efficient evaluation of the Hartree-Fock exact exchange (HFX) in ab initio molecular dynamics simulations, specifically tailored for condensed phase simulations. Our developments allow one to achieve the necessary accuracy for the evaluation of the HFX in a highly controllable manner. We show here that our solutions can take great advantage of the latest trends in HPC platforms, such as extreme threading, short vector instructions and highly dimensional interconnection networks. Indeed, all these trends are evident in the IBM Blue Gene/Q supercomputer. We demonstrate an unprecedented scalability up to 6,291,456 threads (96 BG/Q racks) with a near perfect parallel efficiency, which represents a more than 20-fold improvement as compared to the current state of the art. In terms of reduction of time to solution, we achieved an improvement that can surpass a 10-fold decrease in runtime with respect to directly comparable approaches. We exploit this development to enhance the accuracy of DFT based molecular dynamics by using the PBE0 hybrid functional. This approach allowed us to investigate the chemical behavior of organic solvents in one of the most challenging research topics in energy storage, lithium/air batteries, and to propose alternative solvents with enhanced stability to ensure an appropriate reversible electrochemical reaction. This step is key for the development of a viable lithium/air storage technology, which would have been a daunting computational task using standard methods. Recent research has shown that the electrolyte plays a key role in non-aqueous lithium/air batteries in producing the appropriate reversible electrochemical reduction. In particular, the chemical degradation of propylene carbonate, the typical electrolyte used, by lithium peroxide has been demonstrated by molecular dynamics simulations of highly realistic models. Reaching the necessary high accuracy in these simulations is a daunting computational task using standard methods.