Understanding Complex Systems Symposium

Abstracts

This talk will present an overview of the basic objectives of the field of computer vision, the fundamental problems that make computer vision complex, and the progress made to date.

An outstanding question in social insect biology is how the behavioral decisions of hundreds or thousands of individual workers yields adaptive behavior patterns at the level of the colony. Division of labor is one type of behavior that is a notable feature of colonies and that is believed to be largely responsible for their tremendous ecological success. "Response threshold" models postulate that division of labor results from individual variation among workers in their responsiveness to specific stimuli associated with each task. Threshold models explain how workers can show both task specialization and behavioral flexibility, two notable features of social insect division of labor. One key property of threshold models is that the stimulus for a task, which indicates the level of demand for that task, should remain close to a steady state value that is dependent mainly on the distribution of response thresholds among the workers. This implies that threshold distributions should be adjusted so that a colony can meet its task demands and allocate workers to tasks accordingly; in effect, a colony should actively regulate its levels of task demands. A separate line of reasoning, based on colony life history strategies, suggests that colonies should allocate most of their workers and their labor to "production" tasks, mainly foraging and rearing brood, and as little as possible to "maintenance" tasks, such as cleaning and repairing the nest. Combining the inferences from threshold models and from life history theory leads to the conclusion that most of a colony's labor should be allocated to production tasks, but that the highest priority should be given to maintenance tasks. Preliminary evidence from the behavior of leaf-cutting ants will be presented to support this conclusion.

When a kidney cell is subjected to in-plane oscillations using a magnetic bead, the cell undergoes oscillations at the same frequency in the three direction x,y and z. The oscillation is in the x direction. At a certain critical force (distance of the magnetic tweezer to the cell) the oscillations in the vertical direction start to oscillate at twice the period or half frequency.We will use classical mechanics to investigate this phenomena and show that it is well defined instability and is similar to a liquid being vibrated vertically.

The question addressed is “Can a simple model of the firm using elementary concepts from analysis and physics give both qualitative and quantitative insight for managing a complex business?” The prospects appear optimistic based upon a linear model for demand of the N products competing within a market segment. The model uses a Taylor expansion to represent product demand in terms of the values and prices of the N products and shows that three fundamental metrics – value, cost and the pace of innovation – drive the bottom-line metrics of cash flow and market share. The fundamental metrics are a function of the degree of order within the firm but an algorithm connecting the order parameters to the bottom-line does not exist. As a result, the order parameters must be treated as the control factors in experiments designed to discover pathways for continuous improvement. Applications discussed include assessing the value of proposed product improvements using marketing research, pricing new products and projecting the behavior of the price of a stock over time. Because management’s role is largely limited to managing order, it needs to be skilled in experimental design and analysis.

Circadian rhythms, such as the daily sleep/wake cycle, emerge from recurrent, ordered sets of complex cellular processes within neurons of the mammalian suprachiasmatic nucleus (SCN) of the brain. The driving mechanism issues from interactions of clock genes in a transcriptional/ translational negative feedback loop. Superimposed upon these molecular oscillations, cellular elements are driven in rhythmic, circadian variation. We have parsed this dynamic 24-h process into distinct time domains. Signaling mechanisms that access the clockworks provide insights into these cellular regulatory processes, which include spatiotemporal regulation of Ca+2i and cellular redox state. These studies provide insights into decision-making mechanisms that permit plasticity in clock state.

When foraging, animals weigh the predicted gains and losses of a feeding attempt in terms of nutrient gain, energy expenditure, and risks from noxious prey defense and predation. How is this done? A basic neural network model has been derived from studies of the predatory sea-slug Pleurobranchaea californica. Although simple in brain and body, this animal efficiently integrates factors of sensation, internal state and memories of experience to estimate cost-effectiveness in decision between prey attack and avoidance. Our data indicate that motor and pre-motor networks for feeding and avoidance interact to determine behavioral expression. Hunger state is translated to appetite by fluctuating levels of neuromodulator chemicals, which regulate the activity state and excitability of the feeding network. Network state in turn determines behavioral response thresholds to feeding stimuli. The simple network model is demonstrably capable of efficient foraging with the use of the virtual animal’s affect as primary source of information. Concatenation of several levels of complexity could potentially reproduce most aspects of animal behavior.

A variety of legal liability rules have been analyzed with regard to their economic efficiency in settings in which courts are imperfectly informed. By harnessing information, held privately by the litigants, about the asset under dispute, these rules can certainly furnish improved efficiency, compared with simple property rules. They are not, however, perfectly efficient, inasmuch as they do not guarantee that assets be allocated optimally. A more sophisticated liability rule procedure, involving a generalization of the notion of bidding, is developed and shown to ensure optimal asset allocation. Concrete examples of this procedure, chosen to illustrate both correlated and uncorrelated cases, are discussed. This optimal allocation procedure amounts to the continuous extension of certain multi-stage liability rules. It is shown, however, that the procedure can also be constructed via a game-theoretic approach, involving an application of the revelation principle, which encompasses the continuous extension of multi-stage rules. This work was done in collaboration with Sergey Knysh (UIUC) and Ian Ayres (Yale Law School).

The genetic code is a language for translating genetic information into protein primary sequences. When this language is examined using the tools of a cryptanalyst, it is found that the choice of four bases (A, C, T, G) results in a locally maximal redundancy. I will discuss the relationship between redundancy and genetic coding, and discuss some of the possible evolutionary reasons for this local maximum, including its impact on conservation of amino acid polar requirement.

With the new single molecule techniques, we can detect the structural changes of individual biological molecules over many cycles, thereby determining the transition rates very precisely. Although these molecules are supposed to be identical to each other, each molecule's behavior can be significantly different from others. This molecular individualism, also called memory effect, can be both static and dynamic. We will discuss two specific examples of simple biological systems, DNA four-way junction and RNA folding.

A relatively new technique for the spatial modeling of organic systems has been developed. The process uses iconographic programming to allow a variety of experts to cooperative develop a typical cellular model that all of them can understand and agree accurately captures their collective knowledge. A second program is used to simultaneously place this model in all the cells covering a specified area. This program also interconnects all cells as specified in the cellular model. The parameters of each cellular model are set by a series of maps covering the cellular collection. This process has been applied to the spatial spread of disease, the plight of endangered species and most recently, to the economic and environmental impact of urban sprawl and its control.

Systematic experimental deviations from theoretical predictions derived for water retention characteristics of fractal porous media have previously been interpreted in terms of continuum percolation theory (at low moisture contents, below the critical volume fraction of water, á c, capillary flow ceases). In other work, continuum percolation theory was applied to find the hydraulic conductivity as a function of saturation for saturations high enough to guarantee percolation of capillary flow. Now these two problems are further linked, using percolation theory to estimate non-equilibrium water contents at matric potential values such that the equilibrium water content is too low for percolation of capillary flow paths. Also an analytical scheme is developed to allow an estimation of the value of á c from physical parameters.

full paper

In his classic work Atom and Organism, Walter Elsasser describes "irreducable complexity" as a condition in which actual states are a rare subpopulation of possible states. Biological systems are irreducably complex even at the molecular level, due to the combinatorial explosion of possible ways that a peptide chain could possibly grow, in contrast to the limited (although quite large) number of ways that peptide chains actually grow. Possible motions of folded proteins are similarly uncountably many. Multiscale analysis and simulation comprise a powerful tool for usefully describing and understanding many aspects of irreducably complex systems. This talk will describe multiscale analysis and computation as it is used in our laboratory to understand the details of ion flow through protein channels.

How do atoms arrange themselves and why in multicomponent materials? I will briefly overview (by analogy to phonons) an electronic-based thermodynamic theory to address this question quantitatively. Mainly, I will highlight the complexity of "Chemical Ordering Modes" in multi-component systems and how one can uniquely characterize ordering instabilities. This "new" analysis is also required to understand scattering data from complex material. Example with direct comparison to experimental data is made (e.g., Phil. Mag. Letts, 79, 551 (1999)).

Our understanding of what constitutes a "cancer" is becoming more and more refined. Intricate and complex pathways are involved in the development of an organism. The organism originates as a single cell and progresses through rapid proliferation and differentiation, followed by steady state repair and repletion, and finally senescence. The genetic information for every part of the mature organism exists in the genome of each cell, yet the individual differentiated cell never expresses most of this genetic information. By studying aberrancies in the system, such as the development of neoplasia, mechanisms that control the process of gene activation and repression are being elucidated at a rapid rate. In an article entitled “The Hallmarks of Cancer” (Cell 100:57-70, 2000) Hanahan and Weinberg presented a unifying conceptual model of cancer biology. Based on the past quarter century of intensive research, common themes in cancer have emerged. The “Hanahan-Weinberg Model” identifies 6 global traits common to all cancers. These 6 factors are: self-sufficiency in growth signaling, insensitivity to antigrowth signals, evasion of programmed cell death, development of limitless replicative potential, the capacity for sustained angiogenesis, and upregulation of genes associated with tissue invasion and metastasis. The acquisition of each of these physical characteristics by the cancer cell represents the successful breaching of an evolutionarily anticancer strategy developed by the multicellular organism to maintain physiologic homeostasis. The cancer cell within the living host eukaryote can be considered analogous to a single-celled organism that is in direct competition for scarce resources. Thus, from the perspective of the deranged cell, acquisition of these cancer-associated traits is appropriately adaptive to ensure survival in a stressful environment, but from the perspective of the host the development of a population of cells with these traits is ultimately maladaptive and lethal. While each cancer is unique, with myriad individual mutations and epigenetic changes manifest, it may be possible in the future to target specific anticancer strategies to these common cancer traits. Advances in cancer research will be hastened by the advent of important technological breakthroughs, such as the widespread application of the high throughput capabilities of genomics, proteomics, and pharmacogenomics. The future of cancer medicine may be radically different within the next decade because of these advances in our understanding basic cancer biology.

Although the immune system represents the most thoroughly understood system in mammalian biology, there remains significant hurdles to exploiting this knowledge for therapeutic purposes. The present discussion will attempt to place in context the complexity of the immune system, how cancer cells can evade the system, and how protein engineering offers some hope for redirecting immune cells against cancer and infectious agents.

In the past several decades, physicists have made great strides in understanding how spatial patterns can arise in systems driven far from equilibrium. Of course, many important issues and significant challenges remain. But, with this sense of progress, many researchers began addressing the question of whether the study of pattern formation could help elucidate the formation of structure in biological systems, often called morphogenesis. Of course, living matter is much more complex than non-living. Yet, this talk will hopefully convince you that not only is this physics-based approach possible, but is in fact extremely promising. There are many processes one could choose to discuss; for definiteness, I will focus on the life cycle of the soil amoeba Dictyostelium discoideum. In this organism, starvation triggers a day-long series of transformations that take solitary amoebae and create a cooperative multicellular organism; the process culminates in a plant-like fruiting body containing spore cells specialized for survival in harsh conditions. Ideas from the physics of pattern formation have been used to help explain the wave field used for cell guidance, the streaming of cells into the aggregate and the collective motions seen in multicellular stages. Currently, several groups are working on the single-cell chemotactic response from a similar perspective.

The Virtual Cell is a modular computational framework that permits construction of models, application of numerical solvers to perform simulations, and analysis of simulation results. A key feature of the Virtual Cell is that it permits the incorporation of realistic experimental geometries within full 3D spatial models. The system is designed for cell biologists to aid both the interpretation and the planning of experiments. This talk will describe the status of the project illustrated with several applications to cell biological problems.

Earlier experiments performed in the group of one of us (KN) demonstrate that young adults achieve better control of isometric force production under haptic and/or visual feedback than children. At the same time the observed complexity of the force signal is increasing with improved control. Here we present a generalization of a simple stochastic map model with threshold response (Y.T. Liu, G. Mayer-Kress, K.M. Newell, 1999) which includes multiple time-scales. We will present simulation results that reproduce the above-mentioned transition from large variance/low dimension to small variance/increased dimension. We conclude with a discussion of potential implication for more general issues of motor learning, development, and control.

Networks of molecular processes within living cells are likely to be robust, in that their performance depends largely on their topology and is relatively insensitive to variations in dynamical parameters. Such robustness may be associated with a low probability of a false negative response, P(F-) -- not responding to an appropriate input -- and/or a low probability of a false positive response, P(F+) -- producing output without appropriate input. These probabilities correlate with a basic aspect of topology, the number of reaction steps (N), differently for metabolic and regulatory networks. The matter-processing pathways of metabolism can not give a F+, producing output without the substrates that are their inputs. For metabolism natural selection tends to lower P(F-) by reducing N, as shown for the pentose phosphate pathway and the Krebs citric acid cycle. Regulatory networks, including those that mediate signaling and gene regulation, can give F+ and F- responses. A F+ response may stimulate inappropriate proliferation or death of the cell. Many signaling networks have more than the minimum N needed to transmit signals; the topology of the additional processes tends to reduce P(F+).

P-type silicon substrate, patterned by an oxide, is immersed into an alcohol in which ultrabright silicon nanoparticles of 1 nm in diameter have been dissolved. Under the influence of an electric current, brought about by positive biasing the substrate, particles get drawn/deposited on the substrate, mostly along the conducting paths. SEM images and fluorescent microscopy of the substrate shows three-dimensional growth of near uniform spherical aggregates of 130 nm, and features of tree-like self-assembly. Avoidance of closed loops, and preference for an angle of branching, are among the features observed.

If the basic idea to classify differential equations (based on renromalization group theory) is straightforwardly applied to biological systems, it may suggest a purely developmental taxonomy. After discussing some general problems of classification, possibilities and limitations of the renromalization-group inspired classification will be discussed.

Changing environments imply changing demands for gene expression. Hence, mechanisms for the control of gene expression are ubiquitous in cells. Intense experimental investigation of these mechanisms over a period of decades has revealed a diverse repertoire of molecular designs. These variations in design, which manifest themselves both in the molecular elements and in the circuitry linking the elements, were originally consider to be historical accidents by most molecular biologists. This was not an unreasonable position, since the variations often appeared to be alternative means of accomplishing the same physiological function. However, systematic analyses and carefully controlled comparisons have since led to the discovery of design principles that provide a more fruitful selectionist explanation for many of these alternative designs. These principles allow one to rationalize alternative designs and make predictions for specific physiologies. Here I will consider just one such physiology, inducible catabolic pathways in bacteria, that illustrates three design principles. These involve the molecular mode of gene control, the coupling of elementary gene circuits, and the metabolic connectivity of the network. These features of the design have a profound influence on the dynamics of gene expression. We will examine each of these design principles and show that designs found in nature have been selected for rapid response on both physiological and evolutionary time scales.

Photosynthetic organisms fuel their metabolism with light energy and have developed for this purpose an efficient apparatus for harvesting sunlight, key features of which had been conceptually established long ago. Recently, the atomic structure of a main protein constituent of the apparatus, as it evolved in purple bacteria, has been solved through a combination of modeling, x-ray crystallography and electron microscopy. This permitted the modeling of the entire light harvesting system, a complex nanometric aggregate of transmembrane proteins. We discuss in this chapter how a still ongoing analysis wrestled from an atomic level structure an explanation of the light harvesting function based on quantum physics. The investigations of the light harvesting system of purple bacteria demonstrate particularly clearly the voyage typical for research in biological physics that starts from a simply stated, known function and proceeds through experimental and theoretical investigations carried out at more and more refined levels of molecular reality: first the macromolecular components of the underlying system are identified and their role characterized, e.g., through spectroscopy; then the complex structures of these components are established at atomic resolution and functionally relevant architectural elements are recognized; finally, through refined observation and theoretical analyses of these elements the physical mechanisms exploited by the organism to achieve the cellular function are determined.

full paper

Silicon nanotechnology can now manufacture logic that incorporates more than 43 million Metal-Oxide-Semiconductor Field Effect Transistors (MOSFETs) into a monolithic integrated circuit (IC). Some of these MOSFETs have a gate or control electrode that is only 130nm long with a gate oxide that insulates the control electrode from the current-carrying channel that is as thin as 1.7nm. Moreover, we have recently shown that further miniaturization is practical. We have produced nanometer-scale MOSFETs or nano-transistors with a gate electrode as shorter than 40nm and a gate oxide thinner 1nm. Inexorably, within the next ten years (according to the ITRS roadmap) the electronics industry is expected to integrate over a billion nanotransistors into a ~3-10cm2 area chip, packing about 5-10 nano-transistors/m2. Integration on this scale, along with the facility for nanofabrication, will enabled new types of ICs. For example, we will show that it is now possible to fabricate ICs so small that they could inserted inside a living cell. Since the cell is the key to biology, this chip could provide unprecedented access to it. We will also show how silicon nanofabrication technology can be used to produce nanometer-scale pores (~2nm in diameter) in an ultra-thin glass membrane (~2nm thick) that function like ion channels in the membrane of a living cell. Such devices may ultimately be used in proteomics or for rapid sequencing of minute amounts of DNA to discover the genetic origin of a disease.

Neurons on culture dishes can be confined to grow along paths of proteins patterned with microlithographic techniques, including microcontact printing. By overlaying these paths on surface microelectrode arrays one can stimulate and record neuroelectric activity, opening the possibility to explore the relationship between form and function in small neuronal networks. However, active networks appear to require glia, considerably complicating the investigations. Progress toward solving these problems will be discussed.

Supported by the National Center for Supercomputer Applications (UIUC)

Photosynthesis provides the reduced carbon for plant growth. Crop production models typically use simple empirical algorithms to predict photosynthetic carbon gain and other physiological processes. This approach ignores the wealth of mechanistic information now available and it precludes investigating the impact of manipulating individual steps in the photosynthetic process on crop production. Knowledge of photosynthesis and its link to carbohydrate distribution are sufficient that mechanistically based dynamic models are now feasible. Our long-term aim is to develop a crop model that would provide a work-bench for numerical experimentation of impacts of manipulating photosynthesis, with a primary objective of predicting and adapting production to rising atmospheric carbon dioxide concentration. To achieve this we are developing a dynamic model that scales mechanistically based linked differential equations to describe leaf photosynthesis. The leaf model will be simulated in parallel for different leaves to represent the complex dynamic light and microenvironment of the crop canopy. Light and microenvironment will be simulated with our physical canopy microclimate model (WIMOVAC). Finally this dynamic canopy photosynthesis model will be integrated onto a phenologically driven crop growth model (SoyGro). This is a first step in developing a photosynthesis workbench that would provide: 1) a framework for storing and summarizing the rich informatics of the photosynthetic process across taxa; 2) quantitative description and simulation of the photosynthetic process, with prediction of missing information; and 3) linkage of a carbon production model to real-time measurement of leaf and canopy photosynthesis.

Photonic crystals are materials that allow us to manipulate light in new and unexpected ways. Semiconducting materials played a tremendous role in microelectronics and we expect photonic crystals to revolutionize the world of microphotonics in a similar way. Colloidal self-assembly and multi-beam interference lithography are great tools to build crystals with interesting optical properties. I will review some recent progress towards constructing photonic band-gap materials and switchable 3D Bragg gratings.