TEACHING STUDENTS TO BUILD BIOLOGICAL SIMULATIONS USING AGENT-BASED MODELING

Background: Computational simulation of biological systems is an important and growing part of biological research. The need for biology students to learn to use modeling and simulation has been recognized by multiple scientific organizations, including AAAS and the National Academy of Sciences. Building a simulation of a biological system allows biology students to attain a much deeper understanding of the system, as well as providing in introduction to computational methods. Methods: We have developed a course, Simulation in Biology, which can be easily taught by biology faculty at many different kinds of institutions. Students in Simulation in Biology use the freely available graphical programming language Starlogo to design, construct, and evaluate a simulation of a biological system of interest to them. Starlogo employs agent-based modeling, in which students create biological ‘agents’, which can be molecules, cells, or organisms, depending on the chosen system. Agents are programmed by students to follow behavioral ‘rules’ that are known or hypothesized in the biological literature for that system. Starlogo is easy for novice programmers to learn; students produce a visualization of the spread of an epidemic and a simple ecosystem in the first two weeks of class. Students practice hypothesis testing through performing the ‘simulation cycle’: data gathering, simulation building, prediction, and comparison of simulation results with biological data. The course has been offered three times at WPI, and twice at UMass Boston, including one offering by an ecologist with little previous programming experience. The course was evaluated on both campuses by an independent evaluator, using classroom visits, pre- and post-surveys, analysis of blog entries, and student interviews.  Results: Pre- and post- surveys showed positive movement in nearly every category of learning relevant to the project. In particular, gains were observed in hypothesis design and testing, as well as computational thinking. Comments from blogs and interviews showed that building simulations helped students to understand the complexity of biological systems, in addition to developing skill using computational tools. We are developing a ‘kit’ including tutorials and classroom materials to allow dissemination of the class. Conclusions: Our results suggest that the course provides students with both a deeper understanding of biological systems, and an appreciation of the strengths and caveats of using simulations to model them. Supported by NSF TUES Award DUE -1140672.