This past weekend, we happened to have a scientific conference right on campus. I thought this would be a great opportunity for the students in my behavioral ecology seminar to see what it's like to attend one. The "BEER" conference (Biomathematics and Ecology: Education and Research) is not really about behavioral ecology, but there were plenty of talks that were relevant in some way. Also, I liked the idea of talking about mathematical and computational models and the role they play in behavioral ecology -- Mudd students are not only well prepared in math & computer science, but they also tend to get excited about those things in a way that biology students at most other colleges wouldn't.
So I asked all my students to check out the conference (thanks to the generosity of the organizers, they did not have to pay registration fees). Their assignment was to choose two of the following three to attend:
(1) a keynote or plenary
(2) the poster session
(3) two session talks
On Monday, they all reported back on what they had seen. No one ended up going to the poster session, probably because it was late Friday evening, but they chose a nice variety of different talks to go to. I spent the whole time in a special Social Insects as Complex Systems session, so it was great to hear from them about cool research on other topics, like population ecology, physiology, and education. Below, I'm including their summaries of some of the most interesting things they saw.
(And yes, they loved the acronym. Most of them are seniors, so some of them could have even participated in the special social event, beer tasting at Claremont Craft Ales.)
----------
Claremont Colleges, Oct 10-12 2014
Plenary Speakers
Trachette Jackson, Mathematical Models of Tumor Angiogenesis (Fri 7:00 pm)
Tumor-induced angiogenesis is the formation of new blood vessels induced by a tumor. The occurrence of angiogenesis marks the transition of a tumor from a solid benign tumor to an aggressive growing tumor. Once angiogenesis occurs, tumors become very hard to kill. How can we model the complex process of angiogenesis? Dr. Jackson combined two different models to describe the VEGF-Bcl2-CXCL8 signaling pathway and consequences. The first part of the model describes the extracellular environment of endothelial cells (the cells that start forming new blood vessels). This model assumes a steady-state equilibrium of VEGF concentrations, which is a novel feature of her model. The second part uses a cellular Potts model to describe cellular dynamics. This models the formation of new capillaries on a cell-by-cell basis. She then used her model to make predictions about tumor-induced angiogenesis.
-- K Heath
Sarah Hews: Flipping the Classroom (Sat 8:00 am)
How can we get students more involved in research while helping them develop research questions themselves? Sarah Hews from Hampshire College discussed her experience teaching a Calculus class and a Biocomputational modeling course, where students really needed to understand and identify parameters of equations prior to application on homework problems. Students also begin reading journal papers early in their college career to better understand what the field of research is like. Hews hopes that teachers will not be constrained by teaching only content--but really be more involved in helping students identify what students do not know. The key to successful research students is asking questions and taking the initiative to pursue knowledge further outside of class through other resources like Khan Academy.
-- B Yeh
Session Talks
Terrance Quinn, Contemporary Models in Fish Population Dynamics (Sat 9:15 am)
Dr. Quinn examined the applications of population models to determining fisheries productivity, with Alaskan fisheries as a case study. Understanding the processes that drive productivity of fisheries is critical for their maintenance and the livelihoods of thousands of Americans. Models to estimate the changes in population size have become rather ubiquitous, and are in the process of becoming ever more complex. The most common models involve matrices, such as the Leslie or Lefkovitch matrices, to understand the changes in each age/size class. Basic models make use of data obtained through surveys and fishing to estimate annual survivorship, fecundity, and transition probability. More complex models have come to explicitly incorporate recruitment, immigration/emigration, and other factors. Interest has also been placed on the incorporation of stochasticity and an awareness of the limitations of complex models, as too many parameters result in random walks and errors. Additionally, with multiple large datasets, conflicts may emerge, requiring the use of smaller effective sample sizes for actionable information. Finally, recent work has been investigating the use of Bayesian or quasi-Bayesian methods to parameterize the models. Carrying forward, we can expect to see the implementation of more complexity and stochasticity in models, and correspondingly more sophisticated statistical methods.
-- S Johnson
Richard Watanabe, Leveraging Mathematical Modeling in Human Genetic Studies of Type 2 Diabetes (Sat 10:15 am)
Many factors contribute to the onset of diabetes, including beta-cell failure and failure to produce insulin. Genetics is historically known to be a factor as well, and 80 loci have been associated with the risk of acquiring type-2 diabetes. The majority of the SNPs within these loci are intronic or intergenic, and thus are not found within the coding domain. Therefore, it appears that most SNPs associated with the onset of diabetes via genome-wide association studies (GWAS) do not appear to be so. In response, Richard Watanabe and his lab has been leveraging mathematical models to assess the quality of quantitative phenotypes for association studies. They have given the classic mathematical model for characterizing the glucose-insulin feedback a unique set of parameters related to observable phenotypes that can be related back to association tests. They then use mathematical models for follow-up characterization, simulating the effects of genetic variation on the pathogenicity of type 2 diabetes.
-- K Agwamba
Jennifer Fewell, Scaling of energetics and the organization of work in harvester ant colonies (Sat 1:00 pm)
Dr. Fewell’s lab is interested in ant colonies, their metabolic rates, and how they get their work done. In particular, they wanted to look at how division of labor of ant colonies changes with colony size. Work done by colonies can be thought of in several aspects: total task output, allocation of work, and division of labor. Division of labor in their model was modeled as a quantitative concept. Division of labor = 0 means that all ants that do work, work on all tasks equally. Division of labor = 1 means that each ant only ever does one specialized task. They made mathematical models that could measure on a scale from 0.0-1.0 the division of labor of different colonies. Another part of their model was the set point threshold model. This model is that certain individuals in a colony have a higher sensitivity to certain stimuli, causing them to begin a certain task sooner than other individuals and allowing them to specialize. A good analogy for this is the dirty dishes in the sink - a person who hates clutter and a person who doesn’t even notice clutter live together, and as the sink fills up with dishes, the person who is overly sensitive to clutter will always do the dishes first and the other person will never do the dishes. As colony size grows, they expected there to be greater variance in thresholds for individuals in the ant colony, causing division of labor and specialization to increase. They tested this with ant colonies of many different ages and sizes, and got division of labor from 0.15 to 0.40, increasing with size. They also found that for larger colonies the median speed of the ants was reduced and more workers were stationary, suggesting that the ants were more spatially specialized as well with larger colonies. These findings are relevant to studying efficient ways to get work done in other organisms or even human society.
-- M Hansen
Francisco Valero-Cuevas, Finger Dexterity: Manipulating objects on the edge of stability
(Sat 2:00 pm)
Dr. Valero-Cuevas’ lab focuses on biomechanics and neuromuscular control. In particular, human dexterity in the hands. They ran a number of simple dexterity tests while mapping brain activity with an MRI to determine how this information was parsed and dealt with. The basic structure of these tests involved a subject holding a spring between their fingers and compressing it as much as possible without losing control and shooting away sideways. The results found that holding the spring with about half of the maximum force engaged the brain heavily, while pushing it even further, right to the "edge of stability" resulted in a total disappearance of brain function, a result that was not mimicked when the test was repeated with the same force on a non-compressing wooden dowel instead of a spring. This suggests that the brain function for certain functions of dexterity is being outsourced to lower level, lower latency processors outside of the central nervous system. These findings further showed that men were fundamentally more capable at these tasks than women, and linked hand dexterity to that of the foot as well, suggesting that this capability was at least partially responsible for lower rates of certain types of injuries in males, which has opened up more possibilities for analysis and rehabilitation.
-- A Swafford
Timothy Lucas, An Individual-Based Simulation of Chaparral Vegetation Response to Frequent Wildfire (Sun 9:45 am)
Dr. Lucas created a spatial simulation to model the behavior of plants in the Santa Monica Mountains, an area which has recently had a dramatic increase in frequency of wildfires. When plant coverage of the Santa Monica Mountains decreases drastically, the area is put at risk because it increases the threat of invasion by exotic species and decreases slope stability. Because plant coverage is important, Dr. Lucas decided to create a spatial model of the area, where one can see how much of the area is covered in plants. There are three main life histories of the chaparral shrubs in that area: nonsprouters, which are killed by fire but have seeds that germinate in response to fire; facultative sprouters, which resprout after a fire and can reproduce post wildfire; and obligate sprouters, which resprout after a fire but their seeds are destroyed. Also, each shrub needs about six years before it becomes mature and is able to produce seeds. Dr. Lucas ran his model to project the area’s status in 60 years,using the current rate of fire frequency, where the average time between fires is about six years, and found that coverage of the area was reduced from 75% to 9%, and that the nonsprouter plant went locally extinct because during one interval of fire that was less than 6 years, all of the seedlings died, and there were no seeds to be germinated.
-- G Gadbois
Erin N. Bodine, An Agent-Based Model of Santa Cruz Island Foxes (Sun 11:30 am)
This project is looking into the steep population declines of the Santa Cruz Island fox. The populations of these foxes declined quite dramatically (from a population of 1,500 in 1994 to under 100 foxes in 2004). Due to various efforts, the population is now increasing. Erin Bodine and her collaborators wanted to look into the factors that affected the population decline, and they were curious specifically about the allee effect, in which the fitness of an individual in a population is positively correlated with the population density, though here they were looking at the allee effect at the demographic (as opposed to individual) level. They used an agent-based model to see how different factors impact the population growth rates of the foxes. In the model, the foxes had to find a mate, establish a territory, and then reproduce. They ran the model many times , with different starting population sizes, and without an imposed carrying capacity. They found that a carrying capacity arose naturally due to limited resources (mates, territory, food), and that larger original population sizes were more likely to get to the carrying capacity. The important finding was that negative growth rates were more frequently observed in smaller populations. This was an emergent demonstration of the allee effect in the agent based model.
-- A Ruina
Melanie Moses, Beyond pheromones: evolving efficient, robust and scalable swarm robots that emulate ant foraging behaviors (Sun 12:00 pm)
For ants or robots gathering resources as a group, which strategy is best: (1) remembering where resources are and going back for more, (2) telling others where to find them, or some combination of the two? Moses and her students used genetic algorithms to evolve the best strategy in three different types of simulated environments: (1) all the resources are in clumps (2) all the resources are randomly scattered, and (3) some resources are in clumps, and some are randomly scattered. They found that a combination of memory and communication was the one that worked best overall. When they took these evolved strategies and used them to control real robots “gathering resources” (pretending to pick up QR codes) in an arena, they found that they needed to change the model to include errors in recognizing the resources to get the robots to perform better.
-- M Donaldson-Matasci
Diane Thomson, Linking demographic models to climate drivers (Sun 12:00 pm)
Climate change has entered public consciousness through the popular media, frequently through articles citing models that predict describe climate change’s biological impact. Motivated by increased public scrutiny on these predictive models, Thomson took a critical look at the statistical strengths of various strategies. Informed by her background in field research, Thomson considered what choice of demographic and climatic variables were easily measured as well as informative. She ultimately recommended generalized linear mixed models (GLMM) as a robust method for selecting climatic variables to inform demographic models.
-- E Gibson
No comments:
Post a Comment