Scalable biclustering - the future of big data exploration?

Biclustering is a technique of discovering local similarities within data. For many years the complexity of the methods and parallelization issues limited its application to big data problems. With the development of novel scalable methods, biclustering has finally started to close this gap. In this paper we discuss the caveats of biclustering and present its current challenges and guidelines for practitioners. We also try to explain why biclustering may soon become one of the standards for big data analytics.

A new concept for risk analysis relating to the degradation of water reservoirs.

This paper presents a proposal for a procedure by which to analyse the risk of reservoirs being degraded. The body of water assessed for its susceptibility to degradation in line with the proposed procedure is Myczkowce Reservoir, SE Poland. This reservoir has a maximum capacity of ten million m3 and helps provide hydropower, by serving as a surge tank located above the main Solina Reservoir. On the basis of an assessment of its morphometric and hydrological parameters, Myczkowce Reservoir was assigned to the low-resilience category where risk of degradation was concerned. The primary factors responsible for that are limited capacity in relation to shoreline length, a lack of thermal stratification, and a high value for the Schindler index. These and other environmental parameters provided for Myczkowce's assignment to the category of susceptible to the impact of matter supplied by its catchment, with this reflecting the instantaneous nature of the basin, high values for the Ohle coefficient, average catchment slope, and the lack of a septic system. The designated risk level supported Myczkowce's assignment to a category characterised by an "unacceptable" risk of degradation. The proposed method taking two parameters (resilience and susceptibility) into account represents the first universal method for assessing reservoirs without reference to risks such as drought, flooding, or lack of water supply for human consumption. The risk depends only on the reservoir and catchment parameters.

Modeling fibrin aggregation in blood flow with discrete-particles.

Excessive clotting can cause bleeding over a vast capillary area. We study the mesoscopic dynamics of clotting by using the fluid particle model. We assume that the plasma consists of fluid particles containing fibrin monomers, while the red blood cells and capillary walls are modeled with elastic mesh of "solid" particles. The fluid particles interact with each other with a short-ranged, repulsive dissipative force. The particles containing fibrin monomers have a dual character. The polymerization of fibrin monomers into hydrated fibrins is modeled by the change of the interactions between fluid particles from repulsive to attractive forces. This process occurs with a probability being an increasing function of the local density. We study the blood flow in microscopic capillary vessels about 100 microm long and with diameters in order of 10 microm. We show that the model of polymerization reflects clearly the role played by fibrins in clotting. Due to the density fluctuations caused the by the high acceleration, the fibrin chains are produced within a very short time (0.5 ms). Fibrin aggregation modifies the rheological properties of blood, slows down the incipient flow, and entraps the red blood cells, thus forming dangerous clots.

Dynamical clustering of red blood cells in capillary vessels.

We have modeled the dynamics of a 3-D system consisting of red blood cells (RBCs), plasma and capillary walls using a discrete-particle approach. The blood cells and capillary walls are composed of a mesh of particles interacting with harmonic forces between nearest neighbors. We employ classical mechanics to mimic the elastic properties of RBCs with a biconcave disk composed of a mesh of spring-like particles. The fluid particle method allows for modeling the plasma as a particle ensemble, where each particle represents a collective unit of fluid, which is defined by its mass, moment of inertia, translational and angular momenta. Realistic behavior of blood cells is modeled by considering RBCs and plasma flowing through capillaries of various shapes. Three types of vessels are employed: a pipe with a choking point, a curved vessel and bifurcating capillaries. There is a strong tendency to produce RBC clusters in capillaries. The choking points and other irregularities in geometry influence both the flow and RBC shapes, considerably increasing the clotting effect. We also discuss other clotting factors coming from the physical properties of blood, such as the viscosity of the plasma and the elasticity of the RBCs. Modeling has been carried out with adequate resolution by using 1 to 10 million particles. Discrete particle simulations open a new pathway for modeling the dynamics of complex, viscoelastic fluids at the microscale, where both liquid and solid phases are treated with discrete particles. Figure A snapshot from fluid particle simulation of RBCs flowing along a curved capillary. The red color corresponds to the highest velocity. We can observe aggregation of RBCs at places with the most stagnant plasma flow.

A discrete-particle model of blood dynamics in capillary vessels.

We investigate the mechanism of aggregation of red blood cells (RBC) in capillary vessels. We use a discrete-particle model in 3D to model the flow of plasma and RBCs within a capillary tube. This model can accurately capture the scales from 0.001 to 100 microm, far below the scales that can be modeled numerically with classical computational fluid dynamics. The flexible viscoelastic red blood cells and the walls of the elastic vessel are made up of solid particles held together by elastic harmonic forces. The plasma is represented by a system of dissipative fluid particles. Modeling has been carried out using 1 to 3 million solid and fluid particles. We have modeled the flow of cells with vastly different shapes, such as normal and "sickle" cells. The two situations involving a straight capillary and a pipe with a choking point have been considered. The cells can coagulate in spite of the absence of adhesive forces in the model. We conclude that aggregation of red blood cells in capillary vessels can be stimulated by depletion forces and hydrodynamic interactions. The cluster of "sickle" cells formed in the choking point of the capillary efficiently decelerates the flow, while normal cells can pass through. These qualitative results from our first numerical results accord well with the laboratory findings.

Mesoscopic dynamics of colloids simulated with dissipative particle dynamics and fluid particle model.

We report results of numerical simulations of complex fluids, using a combination of discrete-particle methods. Our molecular modeling repertoire comprises three simulation techniques: molecular dynamics (MD), dissipative particle dynamics (DPD), and the fluid particle model (FPM). This type of model can depict multi-resolution molecular structures found in complex fluids ranging from single micelle, colloidal crystals, large-scale colloidal aggregates up to the mesoscale processes of hydrodynamical instabilities in the bulk of colloidal suspensions. We can simulate different colloidal structures in which the colloidal beds are of comparable size to the solvent particles. This undertaking is accomplished with a two-level discrete particle model consisting of the MD paradigm with a Lennard-Jones (L-J) type potential for defining the colloidal particle system and DPD or FPM for modeling the solvent. We observe the spontaneous emergence of spherical or rod-like micelles and their crystallization in stable hexagonal or worm-like structures, respectively. The ordered arrays obtained by using the particle model are similar to the 2D colloidal crystals observed in laboratory experiments. The micelle shape and its hydrophobic or hydrophilic character depend on the ratio between the scaling factors of the interactions between colloid-colloid to colloid-solvent. Unlike the miscellar arrays, the colloidal aggregates involve the colloid-solvent interactions prescribed by the DPD forces. Different from the assumption of equilibrium growth, the two-level particle model can display much more realistic molecular physics, which allows for the simulation of aggregation for various types of colloids and solvent liquids over a very broad range of conditions. We discuss the potential prospects of combining MD, DPD, and FPM techniques in a single three-level model. Finally, we present results from large-scale simulation of the Rayleigh-Taylor instability and dispersion of colloidal slab in 2D and 3D. Electronic supplementary material to this paper can be obtained by using the Springer LINK server located at http://dx.doi.org/10.1007/s00894-001-0068-3.