Clusters, clouds, and grids offer access to large scale computational resources at low cost. This is especially appealing to scientific applications that require a very large scale to compete in the research space. However, the resources available across these platforms differ significantly in their availability, hardware, environment, performance, cost of use, and more. This requires the use of elastic applications that can adapt to the resources available at run-time, transparently handling heterogeneity and failures. In this paper, we present case studies of several elastic applications built using the Work Queue programming framework. From this experience, we offer six general guidelines for the design and implementation of elastic applications that run on thousands of processors.

Molecular modeling is a field that traditionally has large computational costs. Until recently, most simulation techniques relied on long trajectories, which inherently have poor scalability. A new class of methods is proposed that requires only a large number of short calculations, and for which minimal communication between computer nodes is required. We considered one of the more accurate variants called Accelerated Weighted Ensemble Dynamics (AWE) and for which distributed computing can be made efficient. We implemented AWE using the Work Queue framework for task management and applied it to an all atom protein model (Fip35 WW domain). We can run with excellent scalability by simultaneously utilizing heterogeneous resources from multiple computing platforms such as clouds (Amazon EC2, Microsoft Azure), dedicated clusters, grids, on multiple architectures (CPU/GPU, 32/64bit), and in a dynamic environment in which processes are regularly added or removed from the pool. This has allowed us to achieve an aggregate sampling rate of over 500 ns/hour. As a comparison, a single process typically achieves 0.1 ns/hour.

Over the past decade, high performance applications have embraced parallel programming and computing models. While parallel computing offers advantages such as good utilization of dedicated hardware resources, it also has several drawbacks such as poor fault-tolerance, scalability, and ability to harness available resources during run-time. The advent of cloud computing presents a viable and promising alternative to parallel computing because of its advantages in offering a distributed computing model. In this work, we establish directives that serve as guidelines for the design and implementation or identification of a suitable cloud computing framework to build or convert a high performance application to run in the cloud. We show that following these directives leads to an elastic implementation that has better scalability, run-time resource adaptability, fault tolerance, and portability across cloud computing platforms, while requiring minimal effort and intervention from the user. We illustrate this by converting an MPI implementation of replica exchange, a parallel tempering molecular dynamics application, to an elastic cloud application using the Work Queue framework that adheres to these directive. We observe better scalability and resource adaptability of this elastic application on multiple platforms, including a homogeneous cluster environment (SGE) and heterogeneous cloud computing environments such as Microsoft Azure and Amazon EC2.

Protein-protein interactions are often mediated by flexible loops that experience conformational dynamics on the microsecond to millisecond time scales. NMR relaxation studies can map these dynamics. However, defining the network of inter-converting conformers that underlie the relaxation data remains generally challenging. Here, we combine NMR relaxation experiments with simulation to visualize networks of inter-converting conformers. We demonstrate our approach with the apo Pin1-WW domain, for which NMR has revealed conformational dynamics of a flexible loop in the millisecond range. We sample and cluster the free energy landscape using Markov State Models (MSM) with major and minor exchange states with high correlation with the NMR relaxation data and low NOE violations. These MSM are hierarchical ensembles of slowly interconverting, metastable macrostates and rapidly interconverting microstates. We found a low population state that consists primarily of holo-like conformations and is a "hub" visited by most pathways between macrostates. These results suggest that conformational equilibria between holo-like and alternative conformers pre-exist in the intrinsic dynamics of apo Pin1-WW. Analysis using MutInf, a mutual information method for quantifying correlated motions, reveals that WW dynamics not only play a role in substrate recognition, but also may help couple the substrate binding site on the WW domain to the one on the catalytic domain. Our work represents an important step towards building networks of inter-converting conformational states and is generally applicable.

In this paper, a novel framework is presented to estimate protein-protein interactions (PPIs) and domain-domain interactions (DDIs) based on a belief propagation estimation method that efficiently computes interaction probabilities. Experimental interactions, domain architecture, and gene ontology (GO) annotations are used to create a factor graph representation of the joint probability distribution of pairwise protein and domain interactions. Bound structures are used asa priorievidence of domain interactions. These structures come from experiments documented in iPfam. The probability distribution contained in the factor graph is then efficiently marginalized with a message passing algorithm called the sum-product algorithm (SPA). This method is compared against two other approaches: maximum-likelihood estimation (MLE) and maximum specificity set cover (MSSC). SPA performs better for simulated scenarios and for inferring high-quality PPI data ofSaccharomyces cerevisiae. This framework can be used to predict potential protein and domain interactions at a genome wide scale and for any organism with identified protein-domain architectures.

UNLABELLED: Cytoprophet is a software tool that allows prediction and visualization of protein and domain interaction networks. It is implemented as a plug-in of Cytoscape, an open source software framework for analysis and visualization of molecular networks. Cytoprophet implements three algorithms that predict new potential physical interactions using the domain composition of proteins and experimental assays. The algorithms for protein and domain interaction inference include maximum likelihood estimation (MLE) using expectation maximization (EM); the set cover approach maximum specificity set cover (MSSC) and the sum-product algorithm (SPA). After accepting an input set of proteins with Uniprot ID/Accession numbers and a selected prediction algorithm, Cytoprophet draws a network of potential interactions with probability scores and GO distances as edge attributes. A network of domain interactions between the domains of the initial protein list can also be generated. Cytoprophet was designed to take advantage of the visual capabilities of Cytoscape and be simple to use. An example of inference in a signaling network of myxobacterium Myxococcus xanthus is presented and available at Cytoprophet's website. AVAILABILITY: http://cytoprophet.cse.nd.edu.

Abstract An additional reference is needed in the article Cytoprophet: a Cytoscape plug-in for protein and domain interaction networks inference. We include this reference as a matter of addendum. Contact: [email protected]

Computationally complex and data intensive atomic scale biomolecular simulation is enabled via processing in network storage (PINS): a novel distributed system framework to overcome bandwidth, compute, storage, and security challenges inherent to the wide area computation and storage grid. High throughput data generation requirements for our scientific target are overcome through novel aggregate bandwidth capabilities. Biomolecular simulation methods are correlated with the client tools, hybrid database/file server (GEMS), computation engine (Condor), virtual file system adapter (Parrot), and local file servers (Chirp). PINS performance is reported for the path sampling of a solvated protein domain requiring over 1000 simulations with total output data generation on the order of 1TB.

One goal of contemporary proteome research is the elucidation of cellular protein interactions. Based on currently available protein-protein interaction and domain data, we introduce a novel method, Maximum Specificity Set Cover (MSSC), for the prediction of protein-protein interactions. In our approach, we map the relationship between interactions of proteins and their corresponding domain architectures to a generalized weighted set cover problem. The application of a greedy algorithm provides sets of domain interactions which explain the presence of protein interactions to the largest degree of specificity. Utilizing domain and protein interaction data of S. cerevisiae, MSSC enables prediction of previously unknown protein interactions, links that are well supported by a high tendency of coexpression and functional homogeneity of the corresponding proteins. Focusing on concrete examples, we show that MSSC reliably predicts protein interactions in well-studied molecular systems, such as the 26S proteasome and RNA polymerase II of S. cerevisiae. We also show that the quality of the predictions is comparable to the Maximum Likelihood Estimation while MSSC is faster. This new algorithm and all data sets used are accessible through a Web portal at http://ppi.cse.nd.edu.

Biomolecular simulations produce more output data than can be managed effectively by traditional computing systems. Researchers need distributed systems that allow the pooling of resources, the sharing of simulation data, and the reliable publication of both tentative and final results. To address this need, we have designed GEMS, a system that enables biomolecular researchers to store, search, and share large scale simulation data. The primary design problem is striking a balance between generosity and gluttony. On one hand, storage providers wish to be generous and share resources with their collaborators. On the other hand, an unchecked data producer can be gluttonous and easily replicate data unnecessarily until it fills all available space. To balance generosity and gluttony, GEMS allows both storage providers and data producers to state and enforce policies on the consumption of storage and the replication of data. By taking advantage of known properties of simulation data, the system is able to distinguish between high value final results that must be preserved and low value intermediate results that can be deleted and regenerated if necessary. We have built a prototype of GEMS on a cluster of workstations and demonstrate its ability to store new data, to replicate within policy limits, and to recover from failures.

Sharing data and storage space in a distributed system remains a difficult task for ordinary users, who are constrained to the fixed abstractions and resources provided by administrators. To remedy this situation, we introduce the concept of a tactical storage system (TSS) that separates storage abstractions from storage resources, leaving users free to create, reconfigure, and destroy abstractions as their needs change. In this paper, we describe how a TSS can provide a variety of filesystem and database abstractions for unmodified applications without requiring special privileges or kernel changes. A TSS provides performance competitive with NFS for single clients and also scales well for multiple servers and multiple clients. A prototype TSS of 120 disks and 6 TB of storage has been deployed at the University of Notre Dame and used for applications in high energy physics and bioinformatics.

We present COMPUCELL3D, a software framework for three-dimensional simulation of morphogenesis in different organisms. COMPUCELL3D employs biologically relevant models for cell clustering, growth, and interaction with chemical fields. COMPUCELL3D uses design patterns for speed, efficient memory management, extensibility, and flexibility to allow an almost unlimited variety of simulations. We have verified COMPUCELL3D by building a model of growth and skeletal pattern formation in the avian (chicken) limb bud. Binaries and source code are available, along with documentation and input files for sample simulations, at http:// compucell.sourceforge.net.

MOTIVATION: CompuCell is a multi-model software framework for simulation of the development of multicellular organisms known as morphogenesis. It models the interaction of the gene regulatory network with generic cellular mechanisms, such as cell adhesion, division, haptotaxis and chemotaxis. A combination of a state automaton with stochastic local rules and a set of differential equations, including subcellular ordinary differential equations and extracellular reaction-diffusion partial differential equations, model gene regulation. This automaton in turn controls the differentiation of the cells, and cell-cell and cell-extracellular matrix interactions that give rise to cell rearrangements and pattern formation, e.g. mesenchymal condensation. The cellular Potts model, a stochastic model that accurately reproduces cell movement and rearrangement, models cell dynamics. All these models couple in a controllable way, resulting in a powerful and flexible computational environment for morphogenesis, which allows for simultaneous incorporation of growth and spatial patterning. RESULTS: We use CompuCell to simulate the formation of the skeletal architecture in the avian limb bud. AVAILABILITY: Binaries and source code for Microsoft Windows, Linux and Solaris are available for download from http://sourceforge.net/projects/compucell/

ProtoMol is a high-performance framework in C++ for rapid prototyping of novel algorithms for molecular dynamics and related applications. Its flexibility is achieved primarily through the use of inheritance and design patterns (object-oriented programming). Performance is obtained by using templates that enable generation of efficient code for sections critical to performance (generic programming). The framework encapsulates important optimizations that can be used by developers, such as parallelism in the force computation. Its design is based on domain analysis of numerical integrators for molecular dynamics (MD) and of fast solvers for the force computation, particularly due to electrostatic interactions. Several new and efficient algorithms are implemented in ProtoMol. Finally, it is shown that ProtoMol's sequential performance is excellent when compared to a leading MD program, and that it scales well for moderate number of processors. Binaries and source codes for Windows, Linux, Solaris, IRIX, HP-UX, and AIX platforms are available under open source license at http://protomol.sourceforge.net.