A comprehensive comparison of random forests and support vector machines for microarray-based cancer classification

Background  

Cancer diagnosis and clinical outcome prediction are among the most important emerging applications of gene expression microarray technology with several molecular signatures on their way toward clinical deployment. Use of the most accurate classification algorithms available for microarray gene expression data is a critical ingredient in order to develop the best possible molecular signatures for patient care. As suggested by a large body of literature to date, support vector machines can be considered "best of class" algorithms for classification of such data. Recent work, however, suggests that random forest classifiers may outperform support vector machines in this domain.     

Self-Programming Machines (II): Network of Self-Programming Machines Driving an Ashby Homeostat

The progress in artificial intelligence enables us to conceive adaptive systems whose characteristics are nearer and nearer to those of living beings. These characteristics though depend on ingenious choices by the designer of these systems: Initial conditions, parameters, optimisation functions, gradient and measure of fitness within the environment. Nevertheless, in living systems which are non-finalist, there are no programmers or designers to conceive of such ingenious choices. Our paper "Self-Programming Machines (I)" presents a non-finalist model since initial states and functions are randomly chosen at the beginning and once and for all. In spite of the fact that they are non-finalist, these machines always stabilise at fixed points when they are connected to an external process. This paper studies the dynamics of a mono-layered network of self-programming machines driving a real device. "the Ashby homeostat", and shows the striking properties of such networks. This system stabilises only at fixed points even if it is subjected to small perturbations or intentional breakdowns such as a reversal of power supply or disconnection of one or several motors. Real and simulated experiences are compared and theoretical results are demonstrated.     

How a neutral evolutionary ratchet can build cellular complexity

Complex cellular machines and processes are commonly believed to be products of selection, and it is typically understood to be the job of evolutionary biologists to show how selective advantage can account for each step in their origin and subsequent growth in complexity. Here, we describe how complex machines might instead evolve in the absence of positive selection through a process of "presuppression," first termed constructive neutral evolution (CNE) more than a decade ago. If an autonomously functioning cellular component acquires mutations that make it dependent for function on another, pre-existing component or process, and if there are multiple ways in which such dependence may arise, then dependence inevitably will arise and reversal to independence is unlikely. Thus, CNE is a unidirectional evolutionary ratchet leading to complexity, if complexity is equated with the number of components or steps necessary to carry out a cellular process. CNE can explain "functions" that seem to make little sense in terms of cellular economy, like RNA editing or splicing, but it may also contribute to the complexity of machines with clear benefit to the cell, like the ribosome, and to organismal complexity overall. We suggest that CNE-based evolutionary scenarios are in these and other cases less forced than the selectionist or adaptationist narratives that are generally told.     

Neuroinflammatory processes are important in neurodegenerative diseases: an hypothesis to explain the increased formation of reactive oxygen and nitrogen species as major factors involved in neurodegenerative disease development.

The hypothesis, as stated in the title, has arisen from the failure of simpler notions to explain a series of otherwise difficult to understand observations and the mounting evidence, in a broader sense, that inflammatory processes in the CNS are important etiologically in neurodegenerative diseases. Novel aspects include the primacy of inflammatory processes, within the CNS, which leads to increased formation of "proinflammatory" cytokines that lead to increased formation of reactive oxygen species (ROS) and mediation of the upregulation of genes that produce toxic products such as reactive nitrogen species (RNS). Here I utilize important background reports and synthesize ideas to help account for the noted increases in ROS and RNS and their biological reaction products in neurodegenerative diseases. The uniqueness of the CNS inflammatory processes include minimal damping of amplification processes, such as proinflammatory cytokine-mediated cascades, combined with unique genetic defects, that act in combination with other risk factors to repeatedly "spark" the inflammatory cascades to account for some of the major differences in neurodegenerative diseases. This hypothesis can be experimentally examined by development of definitive methods to quantitate unique products that are formed by processes predicted to occur under neurodegenerative conditions.     

Single-molecule imaging and manipulation of biomolecular machines and systems

Background

Biological molecular machines support various activities and behaviors of cells, such as energy production, signal transduction, growth, differentiation, and migration.

Very different performance of the power Doppler modalities of several ultrasound machines ascertained by a microvessel flow phantom

Introduction

In many patients with rheumatoid arthritis (RA) subclinical disease activity can be detected with ultrasound (US), especially using power Doppler US (PDUS). However, PDUS may be highly dependent on the type of machine. This could create problems both in clinical trials and in daily clinical practice. To clarify how the PDUS signal differs between machines we created a microvessel flow phantom.

Methods

The flow phantom contained three microvessels (150, 1000, 2000 microns). A syringe pump was used to generate flows. Five US machines were used. Settings were optimised to assess the lowest detectable flow for each US machine.

Results

The minimal detectable flow velocities showed very large differences between the machines. Only two of the machines may be able to detect the very low flows in the capillaries of inflamed joints. There was no clear relation with price. One of the lower-end machines actually performed best in all three vessel sizes.

Conclusions

We created a flow phantom to test the sensitivity of US machines to very low flows in small vessels. The sensitivity of the power Doppler modalities of 5 different machines was very different. The differences found between the machines are probably caused by fundamental differences in processing of the PD signal or internal settings inaccessible to users. Machines considered for PDUS assessment of RA patients should be tested using a flow phantom similar to ours. Within studies, only a single machine type should be used.     

A dynamic block device reconfiguration algorithm in virtual MapReduce cluster

With the advances of cloud computing and virtualization technologies, running MapReduce applications over clouds has been attracting more and more attention in recent years. However, as a fundamental problem, the performance of MapReduce applications can sometimes be severely degraded due to the overheads from I/O virtualization and resource competitions among virtual machines. In this paper, we propose a dynamic block device reconfiguration algorithm in virtual MapReduce clusters, which reduces the data transfer time between virtual machines and thereby improving the performance of MapReduce applications on top of the clouds. The proposed algorithm utilizes a block device reconfiguration scheme, where a block device attached to a virtual machine can be dynamically detached and reattached to other virtual machines at runtime. This scheme allows us to move files easily across different virtual machines without any network transfers between virtual machines. This algorithm is also dynamic in a sense that it estimates the total data transfer times between virtual machines using multiple regression analysis based on CPU utilization and data size, and adaptively determines a least-cost data transfer path between a mapper virtual machine and a reducer virtual machine. We have implemented our algorithm in Hadoop MapReduce. The benchmarking results showed that the overheads incurred by transferring data from mapper virtual machines to reducer virtual machines are minimized and the execution times of MapReduce applications are shortened up to 14 %.     

Label-free detection methods for protein microarrays

With the growth of the "-omics" such as functional genomics and proteomics, one of the foremost challenges in biotechnologies has become the development of novel methods to monitor biological process and acquire the information of biomolecular interactions in a systematic manner. To fully understand the roles of newly discovered genes or proteins, it is necessary to elucidate the functions of these molecules in their interaction network. Microarray technology is becoming the method of choice for such a task. Although protein microarray can provide a high throughput analytical platform for protein profiling and protein-protein interaction, most of the current reports are limited to labeled detection using fluorescence or radioisotope techniques. These limitations deflate the potential of the method and prevent the technology from being adapted in a broader range of proteomics applications. In recent years, label-free analytical approaches have gone through intensified development and have been coupled successfully with protein microarray. In many examples of label-free study, the microarray has not only offered the high throughput detection in real time, but also provided kinetics information as well as in situ identification. This article reviews the most significant label-free detection methods for microarray technology, including surface plasmon resonance imaging, atomic force microscope, electrochemical impedance spectroscopy and MS and their applications in proteomics research.     

Rapid isolation and single-molecule analysis of ribonucleoproteins from cell lysate by SNAP-SiMPull

Large macromolecular complexes such as the spliceosomal small nuclear ribonucleoproteins (snRNPs) play a variety of roles within the cell. Despite their biological importance, biochemical studies of snRNPs and other machines are often thwarted by practical difficulties in the isolation of sufficient amounts of material. Studies of the snRNPs as well as other macromolecular machines would be greatly facilitated by new approaches that enable their isolation and biochemical characterization. One such approach is single-molecule pull-down (SiMPull) that combines in situ immunopurification of complexes from cell lysates with subsequent single-molecule fluorescence microscopy experiments. We report the development of a new method, called SNAP-SiMPull, that can readily be applied to studies of splicing factors and snRNPs isolated from whole-cell lysates. SNAP-SiMPull overcomes many of the limitations imposed by conventional SiMPull strategies that rely on fluorescent proteins. We have used SNAP-SiMPull to study the yeast branchpoint bridging protein (BBP) as well as the U1 and U6 snRNPs. SNAP-SiMPull will likely find broad use for rapidly isolating complex cellular machines for single-molecule fluorescence colocalization experiments.     

Institutionalization of Life-Cycle Thinking in the Everyday Discourse of Market Actors

The widespread popularity of life-cycle assessment (LCA) is difficult to understand from the point of view of instrumental decision making by economic agents. Ehrenfeld has argued, in a 1997 issue of this journal, that it is the world-shaping potential of LCA that is more important than its use as a decision-making tool. The present study attempts to explore the institutionalization of this "LCA world view" among ordinary market actors. This is important because environmental policy relies increasingly on market-based initiatives. Cognitive and normative assumptions in authoritative LCA documents are examined as empirical data and compared with data from focus group interviews concerning products and the environment with "ordinary" manufacturers, retailers, and consumers in Finland. These assumptions are (1) the "cradle-to-grave" approach, (2) the view that all products have an environmental impact and can be improved, (3) the relativity of environmental merit, and (4) the way responsibility for environmental burdens is attributed. Relevant affinities, but also differences, are identified. It is argued that life-cycle thinking is not primarily instrumental, but rather is gaining a degree of intrinsic value. The study attempts to establish a broader institutional context in which the popularity of LCA can be understood. From the point of view of this broader context, some future challenges for the development of LCA and life-cycle thinking are suggested.     

DNA structural changes as the basis for a nanomolecular device.

There is currently great interest in the design of nanodevices that are capable of performing movements. Protein molecular machines are abundant in biology but it has recently been proposed that nucleic acids could also act as nanomolecular machines in model systems. Several types of movements have been described with DNA machines: rotation, extension-contraction and "scissor-like" opening and closing. Here we analyze the properties of a simple and robust device composed of a single 21-base-long oligonucleotide which relies on a duplex/quadruplex equilibrium fueled by the sequential addition of DNA single-strands, generating a DNA duplex as a by-product. The interconversion between two well-defined topological states induces a five nanometer two-stroke, linear motor type movement, which is detected by FRET spectroscopy.     

VPM tokens: virtual machine-aware power budgeting in datacenters

Power consumption and cooling overheads are becoming increasingly significant for enterprise datacenters, affecting overall costs and the ability to extend resource capacities. To help mitigate these issues, active power management technologies are being deployed aggressively, including power budgeting, which enables improved power provisioning and can address critical periods when power delivery or cooling capabilities are temporarily reduced. Given the use of virtualization to encapsulate application components into virtual machines (VMs), however, such power management capabilities must address the interplay between budgeting physical resources and the performance of the virtual machines used to run these applications. This paper proposes a set of management components and abstractions for use by software power budgeting policies. The key idea is to manage power from a VM-centric point of view, where the goal is to be aware of global utility tradeoffs between different virtual machines (and their applications) when maintaining power constraints for the physical hardware on which they run. Our approach to VM-aware power budgeting uses multiple distributed managers integrated into the VirtualPower Management (VPM) framework whose actions are coordinated via a new abstraction, termed VPM tokens. An implementation with the Xen hypervisor illustrates technical benefits of VPM tokens that include up to 43% improvements in global utility, highlighting the ability to dynamically improve cluster performance while still meeting power budgets. We also demonstrate how VirtualPower based budgeting technologies can be leveraged to improve datacenter efficiency in the context of cooling infrastructure management.

In the Shadows of Medicine and Modernity: Medical Integration and Secular Histories of Religious Healing in Turkey

Building upon the World Health Organization's recent publication WHO Strategy for Traditional Medicine (WHO 2002), this paper examines the historical position of "traditional medicines" at their intersection with the development and modernization of a biomedically based health care system in Turkey. This paper considers how the historical development of Turkey's health care system, as a prominent site for the articulation of the state's broader modernization project, sustained particular formulations of subjectivity and citizenship that were defined in opposition to a set of cultural practices and modes of religious-political authority represented by "traditional medicines." Consequently, projects and policies seeking to formally integrate "complementary" or "alternative" therapies directly confront this past and the various ways in which it is reenacted in constituting the present.     

Theory of molecular machines. I. Channel capacity of molecular machines

Like macroscopic machines, molecular-sized machines are limited by their material components, their design, and their use of power. One of these limits is the maximum number of states that a machine can choose from. The logarithm to the base 2 of the number of states is defined to be the number of bits of information that the machine could "gain" during its operation. The maximum possible information gain is a function of the energy that a molecular machine dissipates into the surrounding medium (Py), the thermal noise energy which disturbs the machine (Ny) and the number of independently moving parts involved in the operation (dspace): Cy = dspace log2 [( Py + Ny)/Ny] bits per operation. This "machine capacity" is closely related to Shannon's channel capacity for communications systems. An important theorem that Shannon proved for communication channels also applies to molecular machines. With regard to molecular machines, the theorem states that if the amount of information which a machine gains is less than or equal to Cy, then the error rate (frequency of failure) can be made arbitrarily small by using a sufficiently complex coding of the molecular machine's operation. Thus, the capacity of a molecular machine is sharply limited by the dissipation and the thermal noise, but the machine failure rate can be reduced to whatever low level may be required for the organism to survive.     

Distributed Shared Arrays: Portable Shared-Memory Programming Interface for Multiple Computer Systems

This paper describes the design and implementation of a parallel programming environment called Distributed Shared Array (DSA), which provides a shared global array abstract across different machines connected by a network. In DSA, users can define and use global arrays that can be accessed uniformly from any machines in the network. Explicit management of array area allocation, replication, and migration is achieved by explicit calls for array manipulation: defining array regions, reading and writing array regions, synchronization, and control of replication and migration. The DSA is integrated with Grid (Globus) services. This paper also describes the use of our model for gene cluster analysis, multiple alignment and molecular dynamics simulation. In these applications, global arrays are used for storing the distance matrix, alignment matrix and atom coordinates, respectively. Large array areas, which cannot be stored in the memory of individual machines, are made available by the DSA. Scalable performance of DSA was obtained compared to that of conventional parallel programs written in MPI.     

“Simultaneous and Complementary Advances”: Mid-century Expectations of the Interaction of Fisheries Science and Management

The simultaneous development of fisheries science and fisheries management that Ray Beverton and Sydney Holt envisioned in 1957 is argued not to have occurred over the past 40 years. The causes of the many failures of fisheries management have been too frequently a matter of informal debate, and almost never a matter of scientific testing of hypotheses. A broader scope for fisheries science is suggested that would be required as a basis for testing hypotheses about the causes of the failure of simultaneous and complementary advances of the science and management of fisheries. The history of the science and management of the International Whaling Commission is used as an example of the scientific scope required to adequately test hypothesized causes of the failures of fisheries management, and to emphasize the relevance of such testing to evaluating the likely success of current management proposals.     

Time-sharing parallel applications through performance-targeted feedback-controlled real-time scheduling

Most parallel machines, such as clusters, are space-shared in order to isolate batch parallel applications from each other and optimize their performance. However, this leads to low utilization or potentially long waiting times. We propose a self-adaptive approach to time-sharing such machines that provides isolation and allows the execution rate of an application to be tightly controlled by the administrator. Our approach combines a periodic real-time scheduler on each node with a global feedback-based control system that governs the local schedulers. We have developed an online system that implements our approach. The system takes as input a target execution rate for each application, and automatically and continuously adjusts the applications’ real-time schedules to achieve those rates with proportional CPU utilization. Target rates can be dynamically adjusted. Applications are performance-isolated from each other and from other work that is not using our system. We present an extensive evaluation that shows that the system remains stable with low response times, and that our focus on CPU isolation and control does not come at the significant expense of network I/O, disk I/O, or memory isolation.