An Optimizing Code Generator for a Class of Lattice-Boltzmann Computations

The Lattice-Boltzmann method (LBM), a promising new particle-based simulation technique for complex and multiscale fluid flows, has seen tremendous adoption in recent years in computational fluid dynamics. Even with a state-of-the-art LBM solver such as Palabos, a user has to still manually write the program using library-supplied primitives. We propose an automated code generator for a class of LBM computations with the objective to achieve high performance on modern architectures. Few studies have looked at time tiling for LBM codes. We exploit a key similarity between stencils and LBM to enable polyhedral optimizations and in turn time tiling for LBM. We also characterize the performance of LBM with the Roofline performance model. Experimental results for standard LBM simulations like Lid Driven Cavity, Flow Past Cylinder, and Poiseuille Flow show that our scheme consistently outperforms Palabos-on average by up to 3x while running on 16 cores of an Intel Xeon (Sandybridge). We also obtain an improvement of 2.47x on the SPEC LBM benchmark.

Insights into the Functional Roles of N-Terminal and C-Terminal Domains of Helicobacter pylori DprA

DNA processing protein A (DprA) plays a crucial role in the process of natural transformation. This is accomplished through binding and subsequent protection of incoming foreign DNA during the process of internalization. DprA along with Single stranded DNA binding protein A (SsbA) acts as an accessory factor for RecA mediated DNA strand exchange. H. pylori DprA (HpDprA) is divided into an N-terminal domain and a C-terminal domain. In the present study, individual domains of HpDprA have been characterized for their ability to bind single stranded (ssDNA) and double stranded DNA (dsDNA). Oligomeric studies revealed that HpDprA possesses two sites for dimerization which enables HpDprA to form large and tightly packed complexes with ss and dsDNA. While the N-terminal domain was found to be sufficient for binding with ss or ds DNA, C-terminal domain has an important role in the assembly of poly-nucleoprotein complex. Using site directed mutagenesis approach, we show that a pocket comprising positively charged amino acids in the N-terminal domain has an important role in the binding of ss and dsDNA. Together, a functional cross talk between the two domains of HpDprA facilitating the binding and formation of higher order complex with DNA is discussed.

Neutron powder diffraction study of Ba3ZnRu2-xIrxO9 (x=0, 1, 2) with 6H-type perovskite structure

The triple perovskites Ba3ZnRu2-xIrxO9 with x = 0, 1, and 2 are insulating compounds in which Ru(Ir) cations form a dimer state. Polycrystalline samples of these materials were studied using neutron powder diffraction (NPD) at 10 and 295 K. No structural transition nor evidence of long range magnetic order was observed within the investigated temperature range. The results from structural refinements of the NPD data and its polyhedral analysis are presented, and discussed as a function of Ru/Ir content. (C) 2015 Elsevier Masson SAS. All rights reserved.

Relative stabilities of condensed face sharing mono- and di-carboranes: CB20H18 and C2B19H18+

The relative energies of triangular face sharing condensed macro polyhedral carboranes: CB20H18 and C2B19H18+ derived from mono- and di-substitution of carbons in (4) B21H18- is calculated at B3LYP/6-31G* level. The relative energies, H center dot center dot center dot H non-bonding distances, NICS values, topological charge analysis and orbital overlap compatibility connotes the face sharing condensed macro polyhedral mono-carboranes, 8 (4-CB20H18) to be the lowest energy isomer. The di-carba- derivative, (36) 4,4'a-C2B19H18+ with carbons substituted in a different B-12 cage in (4) B21H18- in anti-fashion is the most stable isomer among 28 possibilities. This structure has less non-bonding H center dot center dot center dot H interaction and is in agreement with orbital-overlap compatibility, and these two have the pivotal role in deciding the stability of these clusters. An estimate of the inherent stability of these carboranes is made using near-isodesmic equations which show that CB20H18 (8) is in the realm of the possible. (C) 2015 Elsevier B.V. All rights reserved.

PLUTO plus : Near-Complete Modeling of Affine Transformations for Parallelism and Locality

Affine transformations have proven to be very powerful for loop restructuring due to their ability to model a very wide range of transformations. A single multi-dimensional affine function can represent a long and complex sequence of simpler transformations. Existing affine transformation frameworks like the Pluto algorithm, that include a cost function for modern multicore architectures where coarse-grained parallelism and locality are crucial, consider only a sub-space of transformations to avoid a combinatorial explosion in finding the transformations. The ensuing practical tradeoffs lead to the exclusion of certain useful transformations, in particular, transformation compositions involving loop reversals and loop skewing by negative factors. In this paper, we propose an approach to address this limitation by modeling a much larger space of affine transformations in conjunction with the Pluto algorithm's cost function. We perform an experimental evaluation of both, the effect on compilation time, and performance of generated codes. The evaluation shows that our new framework, Pluto+, provides no degradation in performance in any of the Polybench benchmarks. For Lattice Boltzmann Method (LBM) codes with periodic boundary conditions, it provides a mean speedup of 1.33x over Pluto. We also show that Pluto+ does not increase compile times significantly. Experimental results on Polybench show that Pluto+ increases overall polyhedral source-to-source optimization time only by 15%. In cases where it improves execution time significantly, it increased polyhedral optimization time only by 2.04x.

Functional roles of N-terminal and C-terminal domains in the overall activity of a novel single-stranded DNA binding protein of Deinococcus radiodurans

Single-stranded DNA binding protein (Ssb) of Deinococcus radiodurans comprises N- and C-terminal oligonucleotide/oligosaccharide binding (OB) folds connected by a beta hairpin connector. To assign functional roles to the individual OB folds, we generated three Ssb variants: Ssb(N) (N-terminal without connector), Ssb(NC) (N-terminal with connector) and Ssb(C) (C-terminal), each harboring one OB fold. Both Ssb(N) and Ssb(NC) displayed weak single-stranded DNA (ssDNA) binding activity, compared to the full-length Ssb (Ssb(FL)). The level of ssDNA binding activity displayed by SsbC was intermediate between Ssb(FL) and Ssb(N). Ssb(C) and Ssb(FL) predominantly existed as homo-dimers while Ssb(NC)/Ssb(N) formed different oligomeric forms. In vitro, Ssb(NC) or Ssb(N) formed a binary complex with Ssb(C) that displayed enhanced ssDNA binding activity. Unlike Ssb(FL), Ssb variants were able to differentially modulate topoisomerase-I activity, but failed to stimulate Deinococcal RecA-promoted DNA strand exchange. The results suggest that the C-terminal OB fold is primarily responsible for ssDNA binding. The N-terminal OB fold binds weakly to ssDNA but is involved in multimerization. (C) 2015 The Authors. Published by Elsevier B.V. on behalf of the Federation of European Biochemical Societies. This is an open access article under the CC BY-NC-ND license.

PolyMage: Automatic Optimization for Image Processing Pipelines

This paper presents the design and implementation of PolyMage, a domain-specific language and compiler for image processing pipelines. An image processing pipeline can be viewed as a graph of interconnected stages which process images successively. Each stage typically performs one of point-wise, stencil, reduction or data-dependent operations on image pixels. Individual stages in a pipeline typically exhibit abundant data parallelism that can be exploited with relative ease. However, the stages also require high memory bandwidth preventing effective utilization of parallelism available on modern architectures. For applications that demand high performance, the traditional options are to use optimized libraries like OpenCV or to optimize manually. While using libraries precludes optimization across library routines, manual optimization accounting for both parallelism and locality is very tedious. The focus of our system, PolyMage, is on automatically generating high-performance implementations of image processing pipelines expressed in a high-level declarative language. Our optimization approach primarily relies on the transformation and code generation capabilities of the polyhedral compiler framework. To the best of our knowledge, this is the first model-driven compiler for image processing pipelines that performs complex fusion, tiling, and storage optimization automatically. Experimental results on a modern multicore system show that the performance achieved by our automatic approach is up to 1.81x better than that achieved through manual tuning in Halide, a state-of-the-art language and compiler for image processing pipelines. For a camera raw image processing pipeline, our performance is comparable to that of a hand-tuned implementation.

SMO: An Integrated Approach to Intra-array and Inter-array Storage Optimization

The polyhedral model provides an expressive intermediate representation that is convenient for the analysis and subsequent transformation of affine loop nests. Several heuristics exist for achieving complex program transformations in this model. However, there is also considerable scope to utilize this model to tackle the problem of automatic memory footprint optimization. In this paper, we present a new automatic storage optimization technique which can be used to achieve both intra-array as well as inter-array storage reuse with a pre-determined schedule for the computation. Our approach works by finding statement-wise storage partitioning hyper planes that partition a unified global array space so that values with overlapping live ranges are not mapped to the same partition. Our heuristic is driven by a fourfold objective function which not only minimizes the dimensionality and storage requirements of arrays required for each high-level statement, but also maximizes inter statement storage reuse. The storage mappings obtained using our heuristic can be asymptotically better than those obtained by any existing technique. We implement our technique and demonstrate its practical impact by evaluating its effectiveness on several benchmarks chosen from the domains of image processing, stencil computations, and high-performance computing.

Density control of ZnO nanowires grown using Au-PMMA nanoparticles and their growth behavior.

Au nanoparticles stabilized by poly(methyl methacrylate) (PMMA) were used as a catalyst to grow vertically aligned ZnO nanowires (NWs). The density of ZnO NWs with very uniform diameter was controlled by changing the concentration of Au-PMMA nanoparticles (NPs). The density was in direct proportion to the concentration of Au-PMMA NPs. Furthermore, the growth process of ZnO NWs using Au-PMMA NPs was systematically investigated through comparison with that using Au thin film as a catalyst. Au-PMMA NPs induced polyhedral-shaped bases of ZnO NWs separated from each other, while Au thin film formed a continuous network of bases of ZnO NWs. This approach provides a facile and cost-effective catalyst density control method, allowing us to grow high-quality vertically aligned ZnO NWs suitable for many viable applications.

Recognition of Membrane-Bound Fusion-Peptide/MPER Complexes by the HIV-1 Neutralizing 2F5 Antibody : Implications for Anti-2F5 Immunogenicity

13 p.

Control of gene expression by modulated self-assembly

Numerous transcription factors self-assemble into different order oligomeric species in a way that is actively regulated by the cell. Until now, no general functional role has been identified for this widespread process. Here, we capture the effects of modulated self-assembly in gene expression with a novel quantitative framework. We show that this mechanism provides precision and flexibility, two seemingly antagonistic properties, to the sensing of diverse cellular signals by systems that share common elements present in transcription factors like p53, NF-kappa B, STATs, Oct and RXR. Applied to the nuclear hormone receptor RXR, this framework accurately reproduces a broad range of classical, previously unexplained, sets of gene expression data and corroborates the existence of a precise functional regime with flexible properties that can be controlled both at a genome-wide scale and at the individual promoter level.

Oxidation of Volatile Organic Compounds in Aqueous Solution and at the Air-water Interface of Aqueous Microdroplets

Isoprene (ISO),the most abundant non-methane VOC, is the major contributor to secondary organic aerosols (SOA) formation. The mechanisms involved in such transformation, however, are not fully understood. Current mechanisms, which are based on the oxidation of ISO in the gas-phase, underestimate SOA yields. The heightened awareness that ISO is only partially processed in the gas-phase has turned attention to heterogeneous processes as alternative pathways toward SOA.

During my research project, I investigated the photochemical oxidation of isoprene in bulk water. Below, I will report on the λ > 305 nm photolysis of H₂O₂ in dilute ISO solutions. This process yields C₁₀H₁₅OH species as primary products, whose formation both requires and is inhibited by O₂. Several isomers of C₁₀H₁₅OH were resolved by reverse-phase high-performance liquid chromatography and detected as MH⁺ (m/z = 153) and MH⁺-18 (m/z = 135) signals by electrospray ionization mass spectrometry. This finding is consistent with the addition of ·OH to ISO, followed by HO-ISO· reactions with ISO (in competition with O₂) leading to second generation HO(ISO)₂· radicals that terminate as C₁₀H₁₅OH via β-H abstraction by O₂.

It is not generally realized that chemistry on the surface of water cannot be deduced, extrapolated or translated to those in bulk gas and liquid phases. The water density drops a thousand-fold within a few Angstroms through the gas-liquid interfacial region and therefore hydrophobic VOCs such as ISO will likely remain in these relatively 'dry' interfacial water layers rather than proceed into bulk water. In previous experiments from our laboratory, it was found that gas-phase olefins can be protonated on the surface of pH < 4 water. This phenomenon increases the residence time of gases at the interface, an event that makes them increasingly susceptible to interaction with gaseous atmospheric oxidants such as ozone and hydroxyl radicals.

In order to test this hypothesis, I carried out experiments in which ISO(g) collides with the surface of aqueous microdroplets of various compositions. Herein I report that ISO(g) is oxidized into soluble species via Fenton chemistry on the surface of aqueous Fe(II)Cl₂ solutions simultaneously exposed to H₂O₂(g). Monomer and oligomeric species (ISO)1-8H⁺ were detected via online electrospray ionization mass spectrometry (ESI-MS) on the surface of pH ~ 2 water, and were then oxidized into a suite of products whose combined yields exceed ~ 5% of (ISO)1-8H⁺. MS/MS analysis revealed that products mainly consisted of alcohols, ketones, epoxides and acids. Our experiments demonstrated that olefins in ambient air may be oxidized upon impact on the surface of Fe-containing aqueous acidic media, such as those of typical to tropospheric aerosols.

Related experiments involving the reaction of ISO(g) with ·OH radicals from the photolysis of dissolved H₂O₂ were also carried out to test the surface oxidation of ISO(g) by photolyzing H₂O₂(aq) at 266 nm at various pH. The products were analyzed via online electrospray ionization mass spectrometry. Similar to our Fenton experiments, we detected (ISO)1-7H⁺ at pH < 4, and new m/z⁺ = 271 and m/z^- = 76 products at pH > 5.

Structural and Biophysical Characterization of Variants of the Mechanosensitive Channel of Large Conductance (MSCL)

The ability to sense mechanical force is vital to all organisms to interact with and respond to stimuli in their environment. Mechanosensation is critical to many physiological functions such as the senses of hearing and touch in animals, gravitropism in plants and osmoregulation in bacteria. Of these processes, the best understood at the molecular level involve bacterial mechanosensitive channels. Under hypo-osmotic stress, bacteria are able to alleviate turgor pressure through mechanosensitive channels that gate directly in response to tension in the membrane lipid bilayer. A key participant in this response is the mechanosensitive channel of large conductance (MscL), a non-selective channel with a high conductance of ~3 nS that gates at tensions close to the membrane lytic tension.

It has been appreciated since the original discovery by C. Kung that the small subunit size (~130 to 160 residues) and the high conductance necessitate that MscL forms a homo-oligomeric channel. Over the past 20 years of study, the proposed oligomeric state of MscL has ranged from monomer to hexamer. Oligomeric state has been shown to vary between MscL homologues and is influenced by lipid/detergent environment. In this thesis, we report the creation of a chimera library to systematically survey the correlation between MscL sequence and oligomeric state to identify the sequence determinants of oligomeric state. Our results demonstrate that although there is no combination of sequences uniquely associated with a given oligomeric state (or mixture of oligomeric states), there are significant correlations. In the quest to characterize the oligomeric state of MscL, an exciting discovery was made about the dynamic nature of the MscL complex. We found that in detergent solution, under mild heating conditions (37 °C – 60 °C), subunits of MscL can exchange between complexes, and the dynamics of this process are sensitive to the protein sequence.

Extensive efforts were made to produce high diffraction quality crystals of MscL for the determination of a high resolution X-ray crystal structure of a full length channel. The surface entropy reduction strategy was applied to the design of S. aureus MscL variants and while the strategy appears to have improved the crystallizability of S. aureus MscL, unfortunately the diffraction qualities of these crystals were not significantly improved. MscL chimeras were also screened for crystallization in various solubilization detergents, but also failed to yield high quality crystals.

MscL is a fascinating protein and continues to serve as a model system for the study of the structural and functional properties of mechanosensitive channels. Further characterization of the MscL chimera library will offer more insight into the characteristics of the channel. Of particular interest are the functional characterization of the chimeras and the exploration of the physiological relevance of intercomplex subunit exchange.

Computationally Guided Monomerization of Red Fluorescent Proteins of the Class Anthozoa

Red fluorescent proteins (RFPs) have attracted significant engineering focus because of the promise of near infrared fluorescent proteins, whose light penetrates biological tissue, and which would allow imaging inside of vertebrate animals. The RFP landscape, which numbers ~200 members, is mostly populated by engineered variants of four native RFPs, leaving the vast majority of native RFP biodiversity untouched. This is largely due to the fact that native RFPs are obligate tetramers, limiting their usefulness as fusion proteins. Monomerization has imposed critical costs on these evolved tetramers, however, as it has invariably led to loss of brightness, and often to many other adverse effects on the fluorescent properties of the derived monomeric variants. Here we have attempted to understand why monomerization has taken such a large toll on Anthozoa class RFPs, and to outline a clear strategy for their monomerization. We begin with a structural study of the far-red fluorescence of AQ143, one of the furthest red emitting RFPs. We then try to separate the problem of stable and bright fluorescence from the design of a soluble monomeric β-barrel surface by engineering a hybrid protein (DsRmCh) with an oligomeric parent that had been previously monomerized, DsRed, and a pre-stabilized monomeric core from mCherry. This allows us to use computational design to successfully design a stable, soluble, fluorescent monomer. Next we took HcRed, which is a previously unmonomerized RFP that has far-red fluorescence (λemission = 633 nm) and attempted to monomerize it making use of lessons learned from DsRmCh. We engineered two monomeric proteins by pre-stabilizing HcRed’s core, then monomerizing in stages, making use of computational design and directed evolution techniques such as error-prone mutagenesis and DNA shuffling. We call these proteins mGinger0.1 (λem = 637 nm / Φ = 0.02) and mGinger0.2 (λem = 631 nm Φ = 0.04). They are the furthest red first generation monomeric RFPs ever developed, are significantly thermostabilized, and add diversity to a small field of far-red monomeric FPs. We anticipate that the techniques we describe will be facilitate future RFP monomerization, and that further core optimization of the mGingers may allow significant improvements in brightness.

Imperfection Insensitive Thin Shells

The buckling of axially compressed cylindrical shells and externally pressurized spherical shells is extremely sensitive to even very small geometric imperfections. In practice this issue is addressed by either using overly conservative knockdown factors, while keeping perfect axial or spherical symmetry, or adding closely and equally spaced stiffeners on shell surface. The influence of imperfection-sensitivity is mitigated, but the shells designed from these approaches are either too heavy or very expensive and are still sensitive to imperfections. Despite their drawbacks, these approaches have been used for more than half a century.

This thesis proposes a novel method to design imperfection-insensitive cylindrical shells subject to axial compression. Instead of following the classical paths, focused on axially symmetric or high-order rotationally symmetric cross-sections, the method in this thesis adopts optimal symmetry-breaking wavy cross-sections (wavy shells). The avoidance of imperfection sensitivity is achieved by searching with an evolutionary algorithm for smooth cross-sectional shapes that maximize the minimum among the buckling loads of geometrically perfect and imperfect wavy shells. It is found that the shells designed through this approach can achieve higher critical stresses and knockdown factors than any previously known monocoque cylindrical shells. It is also found that these shells have superior mass efficiency to almost all previously reported stiffened shells.

Experimental studies on a design of composite wavy shell obtained through the proposed method are presented in this thesis. A method of making composite wavy shells and a photogrametry technique of measuring full-field geometric imperfections have been developed. Numerical predictions based on the measured geometric imperfections match remarkably well with the experiments. Experimental results confirm that the wavy shells are not sensitive to imperfections and can carry axial compression with superior mass efficiency.

An efficient computational method for the buckling analysis of corrugated and stiffened cylindrical shells subject to axial compression has been developed in this thesis. This method modifies the traditional Bloch wave method based on the stiffness matrix method of rotationally periodic structures. A highly efficient algorithm has been developed to implement the modified Bloch wave method. This method is applied in buckling analyses of a series of corrugated composite cylindrical shells and a large-scale orthogonally stiffened aluminum cylindrical shell. Numerical examples show that the modified Bloch wave method can achieve very high accuracy and require much less computational time than linear and nonlinear analyses of detailed full finite element models.

This thesis presents parametric studies on a series of externally pressurized pseudo-spherical shells, i.e., polyhedral shells, including icosahedron, geodesic shells, and triambic icosahedra. Several optimization methods have been developed to further improve the performance of pseudo-spherical shells under external pressure. It has been shown that the buckling pressures of the shell designs obtained from the optimizations are much higher than the spherical shells and not sensitive to imperfections.