Search

Progression of computational resources towards exascale computing makes possible simulations of unprecedented accuracy and complexity in the fields of materials and molecular modelling (MMM), allowing high fidelity in silico experiments on complex materials of real technological interest. However, this presents demanding challenges for the software used, especially the exploitation of the huge degree of parallelism available on exascale hardware, and the associated problems of developing effective workflows and data management on such platforms. As part of the UKs ExCALIBUR exascale computing initiative, the UK-led MMM Design and Development Working Group has worked with the broad MMM community to identify a set of high priority application case studies which will drive future exascale software developments. We present an overview of these case studies, categorized by the methodological challenges which will be required to realize them on exascale platforms, and discuss the exascale requirements, software challenges and impact of each application area.

This year 15 interns join us for a 10 week programme between 28th June and 25th September 2021. The projects they are working on are interdisciplinary and include both cutting-edge AI and cutting-edge chemical discovery and demonstrate how and why they are relevant to AI3SD. The projects must be able to demonstrate valuable outputs both with respect to developing student skill and providing impact to AI3SD.The Interns will be required to produce a poster for the AI3SD Summer Project Symposia 1st – 2nd September 2021, and they will take part in our Skills4Scientists programme that will run weekly across July and August whereby the interns will be given training in a range of research, technical and interpersonal skills, alongside informative career-based events.

This video forms part of the Skills4Scientists Series which has been organised as a joint venture between the Artificial Intelligence for Scientific Discovery Network+ (AI3SD) and the Physical Sciences Data-Science Service (PSDS). This series ran over summer 2021 and aims to educate and improve scientists skills in a range of areas including research data management, python, version control, ethics, and career development. This series is primarily aimed at final year undergraduates / early stage PhD students. This video is a compilation of posters presented at the Skills4Scientists Posters & Careers Symposium. These poster presentations are predominantly from summer students involved in the AI3SD 2021 summer internship program. Higher resolution versions of the posters are available on the poster symposium website: https://www.ai3sd.org/s4s-symposium20...Not all poster presenters requested a recording of their talk. The following posters recordings are included in this compilation video. Poster 1 - Nearer the nearsightedness principle: Large-scale quantum chemical calculations – Andras Vekassy (University of Southampton) Poster 3 - Combining Ultrasonic Methods and Machine Learning Techniques to Assess Baked Products Quality – Erhan Gulsen (University of Nottingham) Poster 4 - Interactive Knowledge-Based Solvent Selection Tool – Hewan Zewdu (University of Nottingham)Poster 5 - CV in High Throughput Chemistry – Jamie Longino (University of Strathclyde)Poster 9 - Dewetting in Thin Liquid Films: Using Sparse Optimization to Learn Evolution Equations – Aspen Fenzl (University of Sheffield)Poster 12 - Creating a merged dataset and its exploration with different Machine Learning algorithms – Maximilian Hoffman (Freie Universität of Berlin)Poster 14 - Bayesian optimisation in Chemistry – Rubaiyat Khondaker (University of Cambridge) Poster 15 - A deep neural network for generation of functional organic materials – Rhyan Barrett (University of Warwick) Thank you to our sponsors Optibrium (https://www.optibrium.com/) and Dotmatics (https://www.dotmatics.com/) who supported this event. These poster presentations were live cartooned by ErrantScience (errantscience.com) which is also available on our YouTube Channel. Sections Intro: (0:00) Andras Vekassy - Nearer the nearsighted principle: Large-scale quantum chemical calculations: (0:17) Erhan Gulsen - Combining Ultrasonic Methods and Machine Learning Techniques to Assess Baked Products Quality: (06:11) Hewan Zewdu - Interactive Knowledge-Based Solvent Selection Tool: (12:09) Jamie Longino - CV in High Throughput Chemistry: (16:52) Aspen Fenzl - Dewetting in Thin Liquid Films: Using Sparse Optimization to Learn Evolution Equations: (21:21) Maximillian Hoffman - Creating a merged dataset and its exploration with different machine learning algorithms: (27:31) Rubaiyat Khondaker - Bayesian optimisation in Chemistry: (34:33) Rhyan Barrett - A deep neural network for generation of functional organic materials: (40:06) Further details from this series can be found at: https://www.ai3sd.org/skills4scientists This video is an output from the AI3SD Network+ (Artificial Intelligence and Augmented Intelligence for Automated Investigations for Scientific Discovery) which is funded by EPSRC under Grant Number EP/S000356/1 and PSDS (Physical Sciences Data science Service) which is funded by EPSRC under Grant Number EP/S020357/1.

The ensemble of structures generated by molecular mechanics (MM) simulations is determined by the functional form of the force field employed and its parameterization. For a given functional form, the quality of the parameterization is crucial and will determine how accurately we can compute observable properties from simulations. While accurate force field parameterizations are available for biomolecules, such as proteins or DNA, the parameterization of new molecules, such as drug candidates, is particularly challenging as these may involve functional groups and interactions for which accurate parameters may not be available. Here, in an effort to address this problem, we present ParaMol, a Python package that has a special focus on the parameterization of bonded and nonbonded terms of druglike molecules by fitting to ab initio data. We demonstrate the software by deriving bonded terms’ parameters of three widely known drug molecules, viz. aspirin, caffeine, and a norfloxacin analogue, for which we show that, within the constraints of the functional form, the methodologies implemented in ParaMol are able to derive near-ideal parameters. Additionally, we illustrate the best practices to follow when employing specific parameterization routes. We also determine the sensitivity of different fitting data sets, such as relaxed dihedral scans and configurational ensembles, to the parameterization procedure, and discuss the features of the various weighting methods available to weight configurations. Owing to ParaMol’s capabilities, we propose that this software can be introduced as a routine step in the protocol normally employed to parameterize druglike molecules for MM simulations.

This year 15 interns join us for a 10 week programme between 28th June and 25th September 2021. The projects they are working on are interdisciplinary and include both cutting-edge AI and cutting-edge chemical discovery and demonstrate how and why they are relevant to AI3SD. The projects must be able to demonstrate valuable outputs both with respect to developing student skill and providing impact to AI3SD.The Interns will be required to produce a poster for the AI3SD Summer Project Symposia 1st - 2nd September 2021, and they will take part in our Skills4Scientists programme that will run weekly across July and August whereby the interns will be given training in a range of research, technical and interpersonal skills, alongside informative career-based events.

Innate recognition of viruses is an essential part of the immune response to viral pathogens. This is integral to the maintenance of healthy lungs, which are free from infection and efficient at gaseous exchange. An important component of innate immunity in identifying viruses is the family of C-type collagen containing lectins, also known as collectins. These secreted, soluble proteins are pattern recognition receptors (PRRs) which recognise pathogen-associated molecular patterns (PAMPs), including viral glycoproteins. These innate immune proteins are composed of trimerized units which oligomerise into higher-order structures and facilitate clearance of viral pathogens through multiple mechanisms. Similarly, many viral surface proteins form trimeric configurations, despite not showing primary protein sequence similarities across virus classes and families they belong to. In this review, we discuss the role of lung collectins surfactant proteins A and D (SP-A and SP-D) in viral recognition. Particularly, we focus on the structural similarity and complementarity of these trimeric collectins with the trimeric viral fusion proteins with which, we hypothesise, they have elegantly co-evolved. Recombinant versions of these innate immune proteins may have therapeutic potential in a range of infectious and inflammatory lung diseases including as anti-viral therapeutics.

International audience; Carlo Lamberti opened discussion of the introductory lecture by Bruce Gates: As you have shown, EXAFS plays a crucial role in determining the structure and the nuclearity of nanoparticles (NPs). For each shell, the accuracy of this determination depends on the error bar associated to the coordination number, that strongly correlates with the corresponding Debye–Waller (DW) parameter. This becomes even more important when in situ operando experiments are performed at reaction temperature. Based on your experience, what suggestions can you give to reduce this correlation and increase the potentiality of the technique? Do you believe it is possible to fix or to determine, in a reliable way, DW parameters from independent experimental or computational works? Do you believe that in temperature-dependent experiments it is reliable to adopt th eDebye or the Einstein model to parametrize the evolution of DW parameters?

The solution of the Poisson equation is a crucial step in electronic structure calculations, yielding the electrostatic potential—a key component of the quantum mechanical Hamiltonian. In recent decades, theoretical advances and increases in computer performance have made it possible to simulate the electronic structure of extended systems in complex environments. This requires the solution of more complicated variants of the Poisson equation, featuring nonhomogeneous dielectric permittivities, ionic concentrations with nonlinear dependencies, and diverse boundary conditions. The analytic solutions generally used to solve the Poisson equation in vacuum (or with homogeneous permittivity) are not applicable in these circumstances, and numerical methods must be used. In this work, we present DL_MG, a flexible, scalable, and accurate solver library, developed specifically to tackle the challenges of solving the Poisson equation in modern large-scale electronic structure calculations on parallel computers. Our solver is based on the multigrid approach and uses an iterative high-order defect correction method to improve the accuracy of solutions. Using two chemically relevant model systems, we tested the accuracy and computational performance of DL_MG when solving the generalized Poisson and Poisson–Boltzmann equations, demonstrating excellent agreement with analytic solutions and efficient scaling to ∼10^9 unknowns and 100s of CPU cores. We also applied DL_MG in actual large-scale electronic structure calculations, using the ONETEP linear-scaling electronic structure package to study a 2615 atom protein–ligand complex with routinely available computational resources. In these calculations, the overall execution time with DL_MG was not significantly greater than the time required for calculations using a conventional FFT-based solver.

Tuberculosis is a major global pandemic responsible for more deaths than any other infectious disease, yet no effective vaccine exists. Here, we demonstrate CD1b expression within human tuberculous granulomas, supporting a role for CD1b lipid antigen presentation in host immunity to infection. CD1b presents mycolates, the dominant Mycobacterium tuberculosis (Mtb) cell wall lipid class and key virulence factors, to αβ T cells. We reveal that mycolate tail moieties, distal to the head group, are antigenic determinants for the conserved human germline-encoded mycolyl lipid-reactive (GEM) T cell receptors (TCRs). Computational simulations suggest a putative mechanism whereby lipid-ligand dynamics within CD1b regulate GEM-TCR activity. This work provides insights for the development of major histocompatibility complex (MHC)-independent Mtb lipid vaccines, including those that target GEM T cells.

Field-Effect Transistor sensors (FET-sensors) have been receiving increasing attention for biomolecular sensing over the last two decades due to their potential for ultra-high sensitivity sensing, label-free operation, cost reduction and miniaturisation. Whilst the commercial application of FET-sensors in pH sensing has been realised, their commercial application in biomolecular sensing (termed BioFETs) is hindered by poor understanding of how to optimise device design for highly reproducible operation and high sensitivity. In part, these problems stem from the highly interdisciplinary nature of the problems encountered in this field, in which knowledge of biomolecular-binding kinetics, surface chemistry, electrical double layer physics and electrical engineering is required. In this work, a quantitative analysis and critical review has been performed comparing literature FET-sensor data for pH-sensing with data for sensing of biomolecular streptavidin binding to surface-bound biotin systems. The aim is to provide the first systematic, quantitative comparison of BioFET results for a single biomolecular analyte, specifically Streptavidin, which is the most commonly used model protein in biosensing experiments, and often used as an initial proof-of-concept for new biosensor designs. This novel quantitative and comparative analysis of the surface potential behaviour of a range of devices demonstrated a strong contrast between the trends observed in pH-sensing and those in biomolecule-sensing. Potential explanations are discussed in detail and surface-chemistry optimisation is shown to be a vital component in sensitivity-enhancement. Factors which can influence the response, yet which have not always been fully appreciated, are explored and practical suggestions are provided on how to improve experimental design.

The breast cancer suppressor BRCA2 controls the recombinase RAD51 in the reactions that mediate homologous DNA recombination, an essential cellular process required for the error-free repair of DNA double-stranded breaks. The primary mode of interaction between BRCA2 and RAD51 is through the BRC repeats, which are ∼35 residue peptide motifs that interact directly with RAD51 in vitro. Human BRCA2, like its mammalian orthologues, contains 8 BRC repeats whose sequence and spacing are evolutionarily conserved. Despite their sequence conservation, there is evidence that the different human BRC repeats have distinct capacities to bind RAD51. A previously published crystal structure reports the structural basis of the interaction between human BRC4 and the catalytic core domain of RAD51. However, no structural information is available regarding the binding of the remaining seven BRC repeats to RAD51, nor is it known why the BRC repeats show marked variation in binding affinity to RAD51 despite only subtle sequence variation. To address these issues, we have performed fluorescence polarisation assays to indirectly measure relative binding affinity, and applied computational simulations to interrogate the behaviour of the eight human BRC-RAD51 complexes, as well as a suite of BRC cancer-associated mutations. Our computational approaches encompass a range of techniques designed to link sequence variation with binding free energy. They include MM-PBSA and thermodynamic integration, which are based on classical force fields, and a recently developed approach to computing binding free energies from large-scale quantum mechanical first principles calculations with the linear-scaling density functional code onetep. Our findings not only reveal how sequence variation in the BRC repeats directly affects affinity with RAD51 and provide significant new insights into the control of RAD51 by human BRCA2, but also exemplify a palette of computational and experimental tools for the analysis of protein-protein interactions for chemical biology and molecular therapeutics., Author Summary The atomic scale interactions that occur at the interfaces between proteins are fundamental to all biological processes. One such critical interface is formed between the proteins, human BRCA2 and RAD51. BRCA2 binds to and delivers RAD51 to sites of DNA damage, where RAD51 mediates the error-free repair of double-stranded DNA breaks. Mutations in BRCA2 have been linked to breast cancer predisposition. Therefore, an accurate picture of the interactions between these two proteins is of great importance. BRCA2 interacts with RAD51 via eight “BRC repeats” that are similar, but not identical, in sequence. Due to lack of experimental structural information regarding the binding of seven of the eight BRC repeats to RAD51, it is unknown how subtle sequence variations in the repeats translate to measurable variations in their binding affinity. We have used a range of computational methods, firstly based on classical force fields, and secondly based on first principles quantum mechanical techniques whose computational cost scales linearly with the number of atoms, allowing us to perform calculations on the entire protein complex. This is the first study comparing all eight BRC repeats at the atomic scale and our results provide critical insights into the control of RAD51 by human BRCA2.

Statistical block copolymers, composed of donor (D) and acceptor (A) blocks, are a novel type of material for organic photovoltaics (OPVs) devices. In particular a new series of polymers based on PBTZT-stat-BDTT-8, recently developed by Merck, offers high solubility in different solvents, and a high power conversion efficiency (PCE) in different device architectures. Although it is known that the electronic properties of these materials may be significantly affected by attaching different functional groups on different blocks, it is not fully clear how important the influence of the polymer composition (i.e., the D/A block ratio) is, even if previous studies suggest that this might also have an effect. Therefore, the effect of the polymer composition in terms of HOMO, LUMO energies, and band gap was explored by studying a number of long chain oligomers with more than 1000 atoms each and with different D/A ratios. This study, that is novel both conceptually and methodologically, was made possible by using the linear-scaling reformulation of DFT implemented in the ONETEP code. Our results showed that changing the composition has a significant effect on the electronic structure of statistical copolymers, making this an alternative and suitable strategy to obtain materials with desired properties. Also, a systematic analysis of the effect of a range of different substituents placed in the D and A blocks of PBTZT-stat-BDTT-8 was performed in order to investigate how this class of materials responds to functionalization. We found that it is not possible to know a priori using chemical intuition what kind of influence different types of functional groups may have on these systems, and therefore, computational modeling is essential.

Cluster of differentiation 1c (CD1c)-dependent self-reactive T cells are abundant in human blood, but self-antigens presented by CD1c to the T-cell receptors of these cells are poorly understood. Here we present a crystal structure of CD1c determined at 2.4 Å revealing an extended ligand binding potential of the antigen groove and a substantially different conformation compared with known CD1c structures. Computational simulations exploring different occupancy states of the groove reenacted these different CD1c conformations and suggested cholesteryl esters (CE) and acylated steryl glycosides (ASG) as new ligand classes for CD1c. Confirming this, we show that binding of CE and ASG to CD1c enables the binding of human CD1c self-reactive T-cell receptors. Hence, human CD1c adopts different conformations dependent on ligand occupancy of its groove, with CE and ASG stabilizing CD1c conformations that provide a footprint for binding of CD1c self-reactive T-cell receptors.

The Fourier interpolation of 3D data-sets is a performance critical operation in many fields, including certain forms of image processing and density functional theory (DFT) quantum chemistry codes based on plane wave basis sets, to which this paper is targeted. In this paper we describe three different algorithms for performing this operation built from standard discrete Fourier transform operations, and derive theoretical operation counts. The algorithms compared consist of the most straightforward implementation and two that exploit techniques such as phase-shifts and knowledge of zero padding to reduce computational cost. Through a library implementation (tintl) we explore the performance characteristics of these algorithms and the performance impact of different implementation choices on actual hardware. We present comparisons within the linear-scaling DFT code ONETEP where we replace the existing interpolation implementation with our library implementation configured to choose the most efficient algorithm. Within the ONETEP Fourier interpolation stages, we demonstrate speed-ups of over 1.55×.

We present a computational approach for the simulation of extended x-ray absorption fine structure (EXAFS) spectra of nanoparticles directly from molecular dynamics simulations without fitting any of the structural parameters of the nanoparticle to experimental data. The calculation consists of two stages. First, a molecular dynamics simulation of the nanoparticle is performed and then the EXAFS spectrum is computed from “snapshots” of structures extracted from the simulation. A probability distribution function approach calculated directly from the molecular dynamics simulations is used to ensure a balanced sampling of photoabsorbing atoms and their surrounding scattering atoms while keeping the number of EXAFS calculations that need to be performed to a manageable level. The average spectrum from all configurations and photoabsorbing atoms is computed as an Au L3-edge EXAFS spectrum with the FEFF 8.4 package, which includes the self-consistent calculation of atomic potentials. We validate and apply this approach in simulations of EXAFS spectra of gold nanoparticles with sizes between 20 and 60 Å. We investigate the effect of size, structural anisotropy, and thermal motion on the gold nanoparticle EXAFS spectra and we find that our simulations closely reproduce the experimentally determined spectra.

The study of properties and of processes in materials, frequently hinges upon understanding phenomena which originate at the atomic level. In such cases the accurate description of the interactions between large numbers of atoms is critical and in turn requires the accurate description of the electrons which play a crucial role in the bonding of atoms into molecules, surfaces and solids. This can only be achieved by solving the equations of quantum mechanics. These equations are too complicated to solve exactly; however their solutions can be approximated by computational techniques. The most accurate ? but also most computationally demanding ? are the “ab initio” techniques which do not use any empirical adjustable parameters. Amongst them, the Density Functional Theory (DFT) formulation of quantum mechanics stands out as an excellent compromise between accuracy and computational efficiency. However, the applicability of ab initio techniques is severely limited by poor scaling: the computational effort needed to perform an ab initio calculation increases with (at least) the third power of the number of atoms, N. This cubic-scaling bottleneck limits the number of atoms we can study to a few hundred at most, even on parallel supercomputers. To overcome this length-scale limitation, a number of researchers worldwide have been pioneering the development of a novel class of ab initio methods with linear-scaling or “Order N” (O(N)) computational cost which nevertheless retain the same high level of accuracy as the conventional approaches. While physically motivated, such methods have proved particularly hard to develop as they introduce highly non-trivial localisation constraints. Nevertheless, many major obstacles have been overcome and a number of O(N) methods (SIESTA, CONQUEST, ONETEP, etc.) for ground state DFT calculations on systems with a gap (e.g. molecules, semiconductors and insulators) are now available and have reached a state of maturity that allows them to be used to study ”real” materials. The particular focus of this workshop is therefore to look forward to what can be achieved in the next few years. Our aim is twofold: (1) As O(N) methods are currently extending the applicability of DFT calculations to problems involving biomolecules and nanostructures they are leading to completely new levels of understanding of these systems. This CECAM meeting will give us the opportunity to make an appraisal of such large-scale simulations and their potential to connect more directly to experiments. (2) We also want to examine the options for extending linear-scaling to problems that cannot be treated by ground-state DFT but require other, more complex approaches. These include methods for treating metallic systems, excited states and wavefunction-based theories for including electronic correlation. Finding ways to transform these methods to linear-scaling cost, and hence extent their applicability to the nano-scale, is the next big challenge that the community of developers of large-scale electronic structure methods is beginning to face. We hope that this workshop will stimulate these major new O(N) methodological developments by bringing together the leading groups in the development of O(N) DFT methods with the leading groups in the development of metal and excited-state or wavefunction-based methods. Strong emphasis during the workshop will be given to discussion in order to promote the exchange of ideas between different communities (Physics, Chemistry, Materials Science, Biochemistry) which are all interested in large-scale applications with ab initio accuracy but are approaching them from different perspectives.

The nuclear industry has enormous challenges to address in understanding its waste products and their safe disposal. It is extremely expensive and difficult to work with such waste products. As computational chemistry has made so many advances in the last 30 years, the question arises as to whether it can start to answer some of the basic questions. It was in this context that British Nuclear Fuels plc approached the quantum chemistry group at the University of Cambridge. After initial considerations, it was decided to write an entirely new quantum chemistry package to address these fundamental problems. The MAGIC program has been written to model as accurately as possible the properties of heavy-atom (in particular, actinide) complexes in realistic environments. Major requirements were the need to include relativistic effects for which several investigations have been carried out by quantum chemists in recent years. A severe difficulty is the high angular momentum of the occupied orbitals in the actinides. It was also believed that it was very important to include the effects of electron correlation. Again much progress has been made by quantum chemists with this problem. Therefore this code was written to take into account all these advances in a simple enough way that calculations on realistic systems are possible. The program is the result of a collaboration between British Nuclear Fuels plc and the University of Cambridge. The program has been developed with a view to making the implementation of new ideas as straightforward as possible. Hence, the code has a simple modular structure. Individual modules may of course be combined in a script to run more complicated procedures, such as a self-consistent field (SCF) procedure. The aim of such an approach is to maximize the time spent in the science compared with that spent interfacing with the computer code. For the end user a simple graphical user interface through Cerius# is provided. Standard features of the input may be selected easily from individual menus for each module. It is also possible to access more advanced features. Comprehensive help facilities are available within the interface. Use of the visualization tools helps not only to see the results of calculations on large molecules more clearly, but also to present them in a concise and clear way. The program has been developed on an SG workstation, but it has been extended to run in parallel on a Cray T3E. This paper is the basic paper which describes in detail the philosophy, science and implementation of the MAGIC project. At the end, sample calculations are reported. Furthermore suggestions are made about how this program may, even at this stage, be used to address problems with actinides in the nuclear industry. In order to place the development of the MAGIC project in context and to make adequate recognition of the contribution of others, this article contains considerable material of a review nature, a brief history of the development of quantum chemistry and density function theory, the treatment of core electrons and relativistic effects, the evaluation of integrals, the treatment of solvent effects and the convergence of the SCF iterations. All are written with calculations on actinide complexes as the ultimate goal.