Fully implicit parallel simulation of single neurons

When a multi-compartment neuron is divided into subtrees such that no subtree has more than two connection points to other subtrees, the subtrees can be on different processors and the entire system remains amenable to direct Gaussian elimination with only a modest increase in complexity. Accuracy is the same as with standard Gaussian elimination on a single processor. It is often feasible to divide a 3-D reconstructed neuron model onto a dozen or so processors and experience almost linear speedup. We have also used the method for purposes of load balance in network simulations when some cells are so large that their individual computation time is much longer than the average processor computation time or when there are many more processors than cells. The method is available in the standard distribution of the NEURON simulation program.     

Parallel network simulations with NEURON

The NEURON simulation environment has been extended to support parallel network simulations. Each processor integrates the equations for its subnet over an interval equal to the minimum (interprocessor) presynaptic spike generation to postsynaptic spike delivery connection delay. The performance of three published network models with very different spike patterns exhibits superlinear speedup on Beowulf clusters and demonstrates that spike communication overhead is often less than the benefit of an increased fraction of the entire problem fitting into high speed cache. On the EPFL IBM Blue Gene, almost linear speedup was obtained up to 100 processors. Increasing one model from 500 to 40,000 realistic cells exhibited almost linear speedup on 2000 processors, with an integration time of 9.8 seconds and communication time of 1.3 seconds. The potential for speed-ups of several orders of magnitude makes practical the running of large network simulations that could otherwise not be explored. Action Editor: Alain Destexhe     

A Molecular Dynamics Computer Simulation Performance Comparison of Java Versus C

Abstract

The relative performance of molecular dynamics (MD) computer simulations of fluids written in ANSI C is compared to that achieved by a comparable program written in Java. The performance of the Java program is shown to be dependent upon its runtime environment. The Java Runtime Environment (JRE) from the Java Development Kit (JDK) 1.2 provides a Just-In-Time (JIT) compiler option on Solaris and Windows 95 platforms which decreases the execution time by approximately 4–10× compared to the standard Java interpreter. The compiled Java implementation of the MD computer simulation runs between 30–100% slower. depending on the platform, compared to the equivalent C implementation. The stability of the two simulations, as measured by conservation of energy is shown to be identical to within ～ 1% over 10⁵ time steps.     

Systolic Loop Methods for Molecular Dynamics Simulation,Generalised for Macromolecules

Abstract

Systolic loop programs have been shown to be very efficient for molecular dynamics simulations of liquid systems on distributed memory parallel computers. The original methods address the case where the number of molecules simulated exceeds the number of processors used. Simulations of large flexible molecules often do not meet this condition, requiring the three- and four-body terms used to model chemical bonds within a molecule to be distributed over several processors. This paper discusses how the systolic loop methods may be generalised to accommodate such systems, and describes the implementation of a computer program for simulation of protein dynamics. Performance figures are given for this program running typical simulations on a Meiko Computing Surface using different number of processors.     

A Multistage Load Distribution Strategy for Three-Dimensional Meshes

We study the problem of scheduling a divisible load in a three-dimensional mesh of processors. The objective is to find partition of a load into shares and distribution of load shares among processors which minimize load processing time subject to communication delays involved in sending load from one processor to another. We propose a new scheduling algorithm which distributes load in a sequence of stages across the network, each stage brings load to a set of processors located at the same distance from the load source. A key feature of our solution is that sets of processors receive load in the order of decreasing processing capacities. We call this scheduling strategy Largest Layer First. A theorem about the processing time attained by the algorithm is stated. Performance of the algorithm is compared to earlier results.     

Mutation Load under Vegetative Reproduction and Cytoplasmic Inheritance

For reasons that remain unclear, even multicellular organisms usually originate from a single cell. Here I consider the balance between deleterious mutations and selection against them in a population with obligate vegetative reproduction, when every offspring is initiated by more than one cell of a parent. The mutation load depends on the genomic deleterious mutation rate U, strictness of selection, number of cells which initiate an offspring n, and the relatedness among the initial cells. The load grows with increasing U, n and strictness of selection, and declines when an offspring is initiated by more closely related cells. If Un >> 1, the load under obligate vegetative reproduction may be substantially higher than under sexual or asexual reproduction, which may account for its rarity. In nature obligate vegetative reproduction seems to be more common and long term in taxa whose cytological features ensure a relatively low load under it. The same model also describes the mutation load under two other modes of inheritance: (1) uniparental transmission of organelles and (2) reproduction by division of multinuclear cells, where each daughter cell receives many nuclei. The load declines substantially when the deleterious mutation rate per organelle genome gets lower or when the number of nuclei in a cell sometimes drops. This may explain the small sizes of organelle genomes in sexual lineages and the presence of karyonic cycles in asexual unicellular multinuclear eukaryotes.     

Analysis and computer simulation of accretion patterns in bacterial cultures

Patterned growth of bacteria created by interactions between the cells and moving gradients of nutrients and chemical buffers is observed frequently in laboratory experiments on agar pour plates. This has been investigated by several microbiologists and mathematicians usually focusing on some hysteretic mechanism, such as dependence of cell uptake kinetics on pH. We show here that a simpler mechanism, one based on cell torpor, can explain patterned growth. In particular, we suppose that the cell population comprises two subpopulations —one actively growing and the other inactive. Cells can switch between the two populations depending on the quality of their environment (nutrient availability, pH, etc.) We formulate here a model of this system, derive and analyze numerical schemes for solving it, and present several computer simulations of the system that illustrate various patterns formed. These compare favorably with observed experiments.     

Design and Performance Analysis of Divisible Load Scheduling Strategies on Arbitrary Graphs

In this paper, we consider the problem of scheduling divisible loads on arbitrary graphs with the objective to minimize the total processing time of the entire load submitted for processing. We consider an arbitrary graph network comprising heterogeneous processors interconnected via heterogeneous links in an arbitrary fashion. The divisible load is assumed to originate at any processor in the network. We transform the problem into a multi-level unbalanced tree network and schedule the divisible load. We design systematic procedures to identify and eliminate any redundant processor–link pairs (those pairs whose consideration in scheduling will penalize the performance) and derive an optimal tree structure to obtain an optimal processing time, for a fixed sequence of load distribution. Since the algorithm thrives to determine an equivalent number of processors (resources) that can be used for processing the entire load, we refer to this approach as resource-aware optimal load distribution (RAOLD) algorithm. We extend our study by applying the optimal sequencing theorem proposed for single-level tree networks in the literature for multi-level tree for obtaining an optimal solution. We evaluate the performance for a wide range of arbitrary graphs with varying connectivity probabilities and processor densities. We also study the effect of network scalability and connectivity. We demonstrate the time performance when the point of load origination differs in the network and highlight certain key features that may be useful for algorithm and/or network system designers. We evaluate the time performance with rigorous simulation experiments under different system parameters for the ease of a complete understanding.     

Molecular dynamics simulations of a double unit cell in a protein crystal: volume relaxation at constant pressure and correlation of motions between the two unit cells

Eight molecular dynamics simulations of a double crystal unit cell of ubiquitin were performed to investigate the effects of simulating at constant pressure and of simulating two unit cells compared to a single unit cell. To examine the influence of different simulation conditions, the constant-pressure and constant-volume simulations were each performed with and without counterions and using two different treatments of the long-range electrostatic interactions (lattice-sum and reaction-field methods). The constant-pressure simulations were analyzed in terms of unit cell deformation and accompanying protein deformations. Energetic and structural properties of the proteins in the simulations of the double unit cell were compared to the results of previously reported one-unit-cell simulations. Correlation between the two unit cells was also investigated based on relative translational and rotational movements of the proteins and on dipole fluctuations. The box in the constant-pressure simulations is found to deform slowly to reach convergence only after 5-10 ns. This deformation does not result from a distortion in the structure of the proteins but rather from changes in protein packing within the unit cell. The results of the double-unit-cell simulations are closely similar to the results of the single-unit-cell simulations, and little motional correlation is found between the two unit cells.     

Numerical Simulation of Blood Flow Through Microvascular Capillary Networks

A numerical method is implemented for computing blood flow through a branching microvascular capillary network. The simulations follow the motion of individual red blood cells as they enter the network from an arterial entrance point with a specified tube hematocrit, while simultaneously updating the nodal capillary pressures. Poiseuille’s law is used to describe flow in the capillary segments with an effective viscosity that depends on the number of cells residing inside each segment. The relative apparent viscosity is available from previous computational studies of individual red blood cell motion. Simulations are performed for a tree-like capillary network consisting of bifurcating segments. The results reveal that the probability of directional cell motion at a bifurcation (phase separation) may have an important effect on the statistical measures of the cell residence time and scattering of the tube hematocrit across the network. Blood cells act as regulators of the flow rate through the network branches by increasing the effective viscosity when the flow rate is high and decreasing the effective viscosity when the flow rate is low. Comparison with simulations based on conventional models of blood flow regarded as a continuum indicates that the latter underestimates the variance of the hematocrit across the vascular tree.     

On the Scalability of Dynamic Scheduling Scientific Applications with Adaptive Weighted Factoring

In heterogeneous environments, dynamic scheduling algorithms are a powerful tool towards performance improvement of scientific applications via load balancing. However, these scheduling techniques employ heuristics that require prior knowledge about workload via profiling resulting in higher overhead as problem sizes and number of processors increase. In addition, load imbalance may appear only at run-time, making profiling work tedious and sometimes even obsolete. Recently, the integration of dynamic loop scheduling algorithms into a number of scientific applications has been proven effective. This paper reports on performance improvements obtained by integrating the Adaptive Weighted Factoring, a recently proposed dynamic loop scheduling technique that addresses these concerns, into two scientific applications: computational field simulation on unstructured grids, and N-Body simulations. Reported experimental results confirm the benefits of using this methodology, and emphasize its high potential for future integration into other scientific applications that exhibit substantial performance degradation due to load imbalance.     

Twist limits for late twisting double somersaults on trampoline

An angle-driven computer simulation model of aerial movement was used to determine the maximum amount of twist that could be produced in the second somersault of a double somersault on trampoline using asymmetrical movements of the arms and hips. Lower bounds were placed on the durations of arm and hip angle changes based on performances of a world trampoline champion whose inertia parameters were used in the simulations. The limiting movements were identified as the largest possible odd number of half twists for forward somersaulting takeoffs and even number of half twists for backward takeoffs. Simulations of these two limiting movements were found using simulated annealing optimisation to produce the required amounts of somersault, tilt and twist at landing after a flight time of 2.0 s. Additional optimisations were then run to seek solutions with the arms less adducted during the twisting phase. It was found that 3½ twists could be produced in the second somersault of a forward piked double somersault with arms abducted 8° from full adduction during the twisting phase and that three twists could be produced in the second somersault of a backward straight double somersault with arms fully adducted to the body. These two movements are at the limits of performance for elite trampolinists.     

A computational model of chemotaxis-based cell aggregation

We present a computational model that successfully captures the cell behaviors that play important roles in 2-D cell aggregation. A virtual cell in our model is designed as an independent, discrete unit with a set of parameters and actions. Each cell is defined by its location, size, rates of chemoattractant emission and response, age, life cycle stage, proliferation rate and number of attached cells. All cells are capable of emitting and sensing a chemoattractant chemical, moving, attaching to other cells, dividing, aging and dying. We validated and fine-tuned our in silico model by comparing simulated 24-h aggregation experiments with data derived from in vitro experiments using PC12 pheochromocytoma cells. Quantitative comparisons of the aggregate size distributions from the two experiments are produced using the Earth Mover's Distance (EMD) metric. We compared the two size distributions produced after 24 h of in vitro cell aggregation and the corresponding computer simulated process. Iteratively modifying the model's parameter values and measuring the difference between the in silico and in vitro results allow us to determine the optimal values that produce simulated aggregation outcomes closely matched to the PC12 experiments. Simulation results demonstrate the ability of the model to recreate large-scale aggregation behaviors seen in live cell experiments.     

Control of epidermal stem cell clusters by Notch-mediated lateral induction

Stem cells in the basal layer of human interfollicular epidermis form clusters that can be reconstituted in vitro. In order to supply the interfollicular epidermis with differentiated cells, the size of these clusters must be controlled. Evidence suggests that control is regulated via differentiation of stem cells on the periphery of the clusters. Moreover, there is growing evidence that this regulation is mediated by the Notch signalling pathway. In this paper, we develop theoretical arguments, in conjunction with computer simulations of a model of the basal layer, to show that regulation of differentiation is the most likely mechanism for cluster control. In addition, we show that stem cells must adhere more strongly to each other than they do to differentiated cells. Developing our model further we show that lateral-induction, mediated by the Notch signalling pathway, is a natural mechanism for cluster control. It can not only indicate to cells the size of the cluster they are in and their position within it, but it can also control the cluster size. This can only be achieved by postulating a secondary, cluster wide, differentiation signal, and cells with high Delta expression being deaf to this signal.     

Design and analysis of a distributed grid resource discovery protocol

Computational grids have been emerging as a new paradigm for solving large complex problems over the recent years. The problem space and data set are divided into smaller pieces that are processed in parallel over the grid network and reassembled upon completion. Typically, resources are logged into a resource broker that is somewhat aware of all of the participants available on the grid. The resource broker scheme can be a bottleneck because of the amount of computational power and network bandwidth needed to maintain a fresh view of the grid. In this paper, we propose to place the load of managing the network resource discovery on to the network itself: inside of the routers. In the proposed protocol, the routers contain tables for resources similar to routing tables. These resource tables map IP addresses to the available computing resource values, which are provided through a scoring mechanism. Each resource provider is scored based on the attributes they provide such as the number of processors, processor frequency, amount of memory, hard drive space, and the network bandwidth. The resources are discovered on the grid by the protocol’s discovery packets, which are encapsulated within the TCP/IP packets. The discovery packet visits the routers and look up in the resource tables until a satisfactory resource is found. The protocol is validated by simulations with five different deployment environments.     

Proteomic shifts in embryonic stem cells with gene dose modifications suggest the presence of balancer proteins in protein regulatory networks

Large numbers of protein expression changes are usually observed in mouse models for neurodegenerative diseases, even when only a single gene was mutated in each case. To study the effect of gene dose alterations on the cellular proteome, we carried out a proteomic investigation on murine embryonic stem cells that either overexpressed individual genes or displayed aneuploidy over a genomic region encompassing 14 genes. The number of variant proteins detected per cell line ranged between 70 and 110, and did not correlate with the number of modified genes. In cell lines with single gene mutations, up and down-regulated proteins were always in balance in comparison to parental cell lines regarding number as well as concentration of differentially expressed proteins. In contrast, dose alteration of 14 genes resulted in an unequal number of up and down-regulated proteins, though the balance was kept at the level of protein concentration. We propose that the observed protein changes might partially be explained by a proteomic network response. Hence, we hypothesize the existence of a class of "balancer" proteins within the proteomic network, defined as proteins that buffer or cushion a system, and thus oppose multiple system disturbances. Through database queries and resilience analysis of the protein interaction network, we found that potential balancer proteins are of high cellular abundance, possess a low number of direct interaction partners, and show great allelic variation. Moreover, balancer proteins contribute more heavily to the network entropy, and thus are of high importance in terms of system resilience. We propose that the "elasticity" of the proteomic regulatory network mediated by balancer proteins may compensate for changes that occur under diseased conditions.     

The challenges of understanding glycolipid functions: An open outlook based on molecular simulations

Glycolipids are the most complex lipid type in cell membranes, characterized by a great diversity of different structures and functions. The underlying atomistic/molecular interactions and mechanisms associated with these functions are not well understood. Here we discuss how atomistic and molecular simulations can be used to shed light on the role of glycolipids in membrane structure and dynamics, receptor function, and other phenomena related to emergence of diseases such as Parkinson's. The cases we discuss highlight the challenge to understand how glycolipids function in cell membranes, and the significant added value that one would gain by bridging molecular simulations with experiments. This article is part of a Special Issue entitled Tools to study lipid functions.     

Mathematical Conditions for Induced Cell Differentiation and Trans-differentiation in Adult Cells

We present a theoretical framework for the analysis of the effect of a fully differentiated cell population on a neighboring stem cell population in Multi-Cellular Organisms (MCOs). Such an organism is constituted by a set of different cell populations, each set of which converges to a different cycle from all possible options, of the same Boolean network. Cells communicate via a subset of the nodes called signals. We show that generic dynamic properties of cycles and nodes in random Boolean networks can induce cell differentiation. Specifically we propose algorithms, conditions and methods to examine if a set of signaling nodes enabling these conversions can be found. Surprisingly we find that robust conversions can be obtained even with a very small number of signals. The proposed conversions can occur in multiple spatial organizations and can be used as a model for regeneration in MCOs, where an islet of cells of one type (representing stem cells) is surrounded by cells of another type (representing differentiated cells). The cells at the outer layer of the islet function like progenitor cells (i.e. dividing asymmetrically and differentiating). To the best of our knowledge, this is the first work showing a tissue-like regeneration in MCO simulations based on random Boolean networks. We show that the probability to obtain a conversion decreases with the log of the node number in the network, showing that the model is relevant for large networks as well. We have further checked that the conversions are not trivial, i.e. conversions do not occur due to irregular structures of the Boolean network, and the converting cycle undergoes a respectable change in its behavior. Finally we show that the model can also be applied to a realistic genetic regulatory network, showing that the basic mathematical insight from regular networks holds in more complex experiment-based networks.     

Efficient load balancing on biswapped networks

This paper focuses on devising an efficient algorithm for load balancing on the promising biswapped interconnection networks which were recently proposed as a better architecture over the well-known OTIS networks. The proposed algorithm is called GPM which reduces the number of load balancing steps of the existed algorithms obviously. GPM algorithm first schedules load flows on inter-groups links to achieve the balanced status among groups. Then a general load balancing strategy is executed in each of all groups to balance processor loads. The analytical model proves that GPM algorithm is efficient and results of computer simulation experiment indicate that GPM can implement load balancing in biswapped network interconnected environments efficiently, in terms of various parameters.