1st Place: Tangled Agents: Socio‐Political Power Struggles in Present‐Day Thailand
R. Jordan Crouser, Doctoral Candidate, Computer Science, School of Arts and Sciences
Remco Chang, Assistant Professor, Computer Science, School of Arts and Sciences
Current research in the social and behavioral sciences relies heavily on simulation methods for modeling complex dynamic systems such as political power hierarchies, flock behavior, web analytics, and neural networks. In order to capture the intricate nature of these real‐world phenomena, these models often contain thousands of interacting agents, each with hundred of variables controlling their behaviors and relationships. Because of this massive data generation, there exists a rapidly growing need for better computational and analytical methods for making sense of the results generated by these simulations.
In our research, we start to address this need by developing a new theoretical framework and visualization techniques for analyzing results from large‐scale agent‐based simulations. We begin by framing the analysis of these simulations as a graph (or connectivity) problem. In this graph, a node represents a unique configuration state reached during the simulation, and a directed edge denotes a temporal transition between two states that has been observed during a run of the simulation. When social and political scientists run their agent‐based simulation experiments, the simulation is run a few hundred to a few thousand times, each time with a different random seed. Each of these runs can be thought of as a pathway through the “simulation space”, the set of reachable configurations in the high‐dimensional parameter space. By collecting these runs and collapsing their common states, we are able to reconstruct the overall “shape” of the simulation space.
In this image, we have used this framework to visualize a dataset from a three‐month simulation of political violence in present‐day Thailand. Each node represents a power hierarchy of 31 different socio‐political identities with a single dominant identity. The edge colors represent different runs of the simulation that have been aggregated together. We have organized the graph so that hierarchies that share a dominant identity are clustered together. We are then able to see interesting patterns of socio‐political change; for example, the consistent power struggle between identity 7 (top center) and identity 25 (lower left).
2nd Place: Loop emission from a cosmic string
Ken Olum, Research Associate Professor, Physics and Astronomy, School of Arts and Sciences
Cosmic strings are astronomically long, microphysically thin filaments which may exist in our universe. When strings intersect themselves, they can reconnect and so chop off moving loops, which may be observable sources of cosmic rays or gravitational waves.
The wiggly tube shows the shape of a cosmic string taken from a simulation, at a particular moment in time. This string is about to emit several bursts of loops. The color of the string shows its rate of motion, red being very close to the speed of light and blue being slowest, and the white arrows show the direction and velocity of selected string segments.
Each colored arrow represents the emission of one loop. The starting point of the arrow is the point where the loop breaks off the long string (or a predecessor loop). The width of the arrow is proportional to the size of the loop, and the length and direction of the arrow show the loop velocity. In general the larger loops move more slowly, and some of the smallest loops travel at nearly the speed of light. The color of the loop represents the time of emission: the dark blue loops are just about to be emitted, while the red loops will be emitted last. Starting from the position shown, the string travels toward the right, sweeping across the field of view and emitting several hundred loops as it passes.
3rd Place: Nosoi’s Quilt: Visualizing large comorbidity data
Kenneth Chui, Public Health and Community Medicine, School of Medicine
Comorbidity is the presence of one or more diseases or disorders in addition to the preexisting disease. Knowing the patterns of comorbidity can help establishing possible pathological relationships, and subsequently help medical professionals to better assign a set of diagnostic procedures given a primary disease or disorder. Electronic medical records allow us to investigate the patterns and magnitudes of comorbidity quickly and accurately. This project demonstrates how to visualize large amount of comorbidity data using a modified bubble plot.
The data are hospitalization records of the US elderly aged 65 years old in 2006, provided by the Centers of Medicare and Medicaid Services. Information in each individual‐level record include age, gender, date of admission, and up to 10 diagnosis codes assigned to the patient based on the International Classification of Diseases, 9th edition, clinical modification (ICD‐9‐CM). ICD‐9‐CM has 19 major domains, including up to 1,200 major diseases or disorders. There are about 15 million records in year 2006. This project used only 10% randomly selected cases.
A set of 1,200 binary indicators, each represents one major disease/disorder, was made to capture the presence (indicator = 1) or absence (0) of the condition. The diseases/disorders with a total frequency more than 1,000 (i.e. >10,000 cases in 2006) were retained in the analysis, resulting in 372 major diseases. Pearson’s Correlation tests were then performed to estimate the correlation between each pair of the diseases. The resultant 372×372 correlation matrix was visualized with a bubble plot. Positive correlation coefficients were divided 8 equal groups based on quantiles. To each group a purple color was assigned. Higher saturation indicates more positive correlation. For simplicity and ease of viewing, diseases combinations with negative correlation coefficients were assigned to white.
The squares along the top‐left to bottom‐right indicate high correlations between intra‐domain diagnoses, which are expected. The less prominent rectangles outside the diagonal line indicate comorbidity. Some conditions were also seen to be common, as indicated by long dark vertical or horizontal lines. Techniques such as interactive interface (e.g. mouse‐over information) or data‐driven matrix ranking will help enriching information and pathological insights.
InfoBiology: Scheme showing the preparation and read-out of a SPAM(Steganography by Printed Arrays of Microbes)
Manuel A. Palacios, Chemistry, School of Arts and Sciences
Elena Benito-Peña, Chemistry, School of Arts and Sciences
Mael Manesse, Chemistry, School of Arts and Sciences
David R. Walt, Chemistry, School of Arts and Sciences
In InfoBiology, we use living organisms as carriers of encoded messages. We encode messages by controlling gene expression and/or cell replication and use them as cipher keys for the delivery of information genetically encoded in organisms. The messages, called SPAM (Steganography by Printed Arrays of Microbes) consist of a matrix of fluorescent spots generated by seven strains of E. coli colonies, with each strain expressing a different fluorescent protein. The Scheme presented shows the preparation and read-out of a SPAM. The green arrow follows the sender’s actions to prepare a SPAM, while the red arrow follows the receiver’s action to develop a SPAM. First, broth containing fluorescent bacteria is pipetted into a microtiter plate. Second, a multi-blot pin replicator is used to transfer a small volume of the broth onto a target plate containing the appropriate growth medium. After the undeveloped SPAM is grown it can be transferred to a nitrocellulose or velvet membrane for delivery. The receiver stamps the SPAM onto an appropriate growth medium and develops the SPAM message. Note that the “undeveloped” SPAM does not have a clear color read-out because protein expression has not yet been induced. For illustration, an image of a nitrocellulose membrane containing a “developed” SPAM is shown at bottom left.
Somerville Immigrant Worker Health Project Grant Timeline
David M. Gute, Associate Professor, Civil and Environmental Engineering, School of Arts and Sciences
Justin Kennard, President, Kennard Design
Linda Markarian, Independent Consultant
Alex Pirie, Coordinator, Immigrant Service Providers Group/ Health (ISPG/H)
The Somerville Immigrant Worker Health Project was a collaborative effort lead by a community based organization, Immigrant Service Providers Group/Health (ISPG/H), a health care provider, Cambridge Health Alliance (CHA), and an academic partner, Tufts University, along with other partners representative of the community. These additional partners included; the Haitian Coalition, Community Action Agency of Somerville (CAAS), Brazilian Women’s Group and Massachusetts Coalition for Occupational Safety and Health (MassCOSH) worked together to address occupational health issues for the populations that they serve at the community level as well as to gather quantitative and qualitative information regarding immigrant occupational health. This work began in 2005 and ended in 2010.
Two perspectives shaped the consideration of immigrant occupational health in this project. First, the knowledge that the number of immigrant residents working and living in Somerville is undercounted due to issues of immigrant and legal documentation. Second, the work reported here follows the Environmental Justice model in that we hold to the premise that the environmental and occupational risks borne by immigrant workers are disproportionally distributed in society. Together, these perspectives led us to attempt to reach further into the immigrant community in Somerville while bringing significant resources to the immigrant service agencies with whom we partnered.
The Grant Timeline (pp. 4 and 5) of the attached file represents our distillation and visual presentation of the complexity inherent in this multi-layered work. The work reported on here is representative of Community Based Participatory Research (CBPR).
This approach places a great deal of emphasis on transparency and establishing an equal footing between the community and academic/institution representatives. We believe the use of a Timeline creates a powerful and an easily grasped, by community member and academic partner alike, means of referring to project accomplishments and identifying key milestones. We employed the Timeline as the visual fulcrum of the Program for a Community Meeting held on 10/13/10 at which we reported our preliminary findings and sought comments. We provide pp. 1-3 and 6-8 as context for your consideration of the Timeline. All of the principals contributed. Mr. Kennard is the designer of our submission.
Hierarchical Partitioning of Biochemical Networks into Functional Modules
Gautham Sridharan, Research Assistant, Chemical and Biological Engineering, School of Arts and Sciences
Kyongbum Lee, Associate Professor, Chemical and Biological Engineering, School of Arts and Sciences
Soha Hassoun, Associate Professor, Computer Science, School of Arts and Sciences
The objective of this study was to develop a computational method for the visual analysis of complex networks. One attractive approach is to partition a network into sub-networks, or ‘modules,’ representing clusters of tightly connected or highly interacting components. There is increasing evidence that modular organization is present in many different types of evolved networks ranging from social networks to ecological food webs to biochemical reaction networks. In this analysis, the contribution of feedback loops in determining the modularity of biochemical networks was investigated. In both engineered and evolved systems, feedback confers stability, whereas modularity is thought to dampen the propagation of instability.
The attached figure shows the result of a hierarchical partitioning of ca. 160 major metabolic reactions taking place in the human liver. For interpretation, the modules were compared against ‘textbook’ groupings of reactions based on historical knowledge of biochemistry. Each reaction was assigned to a canonical function group such as glucose metabolism (GLYCO) and mapped to a color (see figure legend). The parent pie chart, or root node (ID: 15900), shows the composition of the overall network, where the relative area of each pie color is proportional to the fraction of reactions belonging to the corresponding canonical function group.
The partition tree shown in the figure was constructed using built-in functions and custom code written in Matlab®. The (x,y) coordinates of the nodes (pie charts) were calculated based on the information regarding the parent and children of each node, which in turn was derived from the partitions. Briefly, the y-coordinates of the pie-charts were obtained as a function of the depth of the node in the hierarchy while the x-coordinates were obtained by calculating the width occupied by each node, which is a function of the number of children passed while traversing to the terminal leaf nodes. The pie charts, which are images represented as mXnX3 matrices storing Red Green Blue intensities, can be placed in a larger matrix according to their position coordinates. A function was written to draw a line between the centers of two pie charts without crossing the pie charts. Matlab converts this larger matrix containing the information for all the pie charts into the attached *.png file. A notable result of the partition shown here is that the nodes become more homogeneous with increased depth.
Illuminating the Resident Referral Process at an Academic Medical Center
Elisabeth E. Bennett, Assistant Professor, Medicine, Tufts University School of Medicine, and Director of Education, Baystate Health
Purpose and Method. The purpose of this project was to illuminate a complex referral process for residents in difficulty at an academic medical center using a qualitative conceptual development method. Stakeholders from the employee assistance program (EAP), human resource consulting (HRC), academic affairs (registrar and designated institutional official), employee health services (EHS), and a representative program director engaged in a conceptual development process facilitated by a qualitative research expert. The resulting graphic illuminated two ways in which residents entered the referral system, which are a) program director referrals and b) self-referrals. The graphic addresses sub-processes, communication patterns, and major responsibilities of stakeholders.
Significance: Residents are medical doctors in graduate medical education programs. They are both employees and students. In addition to clinical and educational work, residents spend up to 20% of their time teaching undergraduate medical students. As such they represent instructional personnel for medical schools. Balancing multiple roles can be difficult given that there are separate laws that govern resident processes, such as FERPA and EEOC. Because patient lives are at stake, program leaders have a critical duty to refer residents who are impaired (e.g. suspected substance abuse, high stress) or if they are having trouble with learning issues (e.g. potential undiagnosed learning disabilities, cognitive processing problems) by describing problem behaviors to referral centers without attempting to diagnosis the cause.
Outcomes: This project illuminated a complex system of referrals. It led to a couple of discoveries, including a) the development of a shared understanding amongst the stakeholders that “return to work under” under the fitness for duty policy should be distinguished from “return to education” since a resident may be able to resume studies before being ready to resume clinical duties after a break from employment and b) the need to identify essential job functions and how these are related to program curricula. Additionally, this project identified an important dividing line between institutional interests and privacy issues when enacting a newly created referral system for learning issues and potential areas of conflicting interests.
Mitochondrial Role in Apoptosi
Fadi Ramadan, Assistant Professor, Medicine, School of Medicine
Apoptosis, or programmed cell death, is a form of “cell suicide” that plays a key role in senescence. During apoptosis, the cell body shrinks, and chromosomal DNA becomes condensed. Eventually, the cell breaks up into several smaller bodies that are engulfed and destroyed by scavenging cells.
Caspases, a family of proteins, are one of the main effectors of apoptosis. The mitochondria are key regulators of the caspase cascade and apoptosis. Release of cytochrome C from mitochondria can lead to the activation of caspases. This effect is mediated through the formation of an apoptosome, a multi-protein complex that contains procaspase 9 among other proteins.
The bcl-2 proteins are a family of proteins involved in the response to apoptosis. Some of these proteins (such as bcl-2 and bcl-XL) are anti-apoptotic, while others (such as Bad or Bax) are pro-apoptotic. The sensitivity of cells to apoptotic stimuli can depend on the balance of pro- and anti-apoptotic bcl-2 proteins. When there is an excess of pro-apoptotic proteins the cells are more sensitive to apoptosis, when there is an excess of anti-apoptotic proteins the cells will tend to be less sensitive.
The pro-apoptotic bcl-2 proteins (found in cytosol) act as sensors of cellular damage or stress. Following cellular stress they relocate to the surface of the mitochondria where the anti-apoptotic proteins are located. This interaction between pro- and anti-apoptotic proteins disrupts the normal function of the anti-apoptotic bcl-2 proteins and can lead to the formation of pores in the mitochondria and the release of cytochrome C and other pro-apoptotic molecules. This in turn leads to formation of the apoptosome (multi-protein complex consisting of cytochrome C, Apaf-1, procaspase 9 and ATP) and activation of the caspase cascade.
The attached illustration demonstrates mitochondrial role in activation of procaspase-9 and role of pro-apoptotic proteins (Bad & Bax) in mitochondrial pore formation and release of cytochrome C (two key steps in apoptosis).