Introduction •

400mM Tris, 20mM acetic acid, 1mM EDTA

40mM Tris, 200mM acetic acid, 1mM EDTA

Water only

The Gelbox Interface •

Commands & Shortcuts •

Discussion •

We built Gelbox to be an interactive tool that could complement the ad hoc learning methods routinely used in laboratory settings. While working with real gels provides valuable hands-on experience, the learning process is constrained by a slow feedback loop. Several hours can pass between making a mistake in preparing a gel and observing the effects of that mistake. Fortunately, computer simulation can reduce the time to fractions of a second between modifying a gel parameter and observing the effects, making cause-and-effect relationships more visible and explicit.

While other gel simulators do exist, they provide only a single static graphical representation of predicted DNA fragment mobility under ideal experimental conditions^8–11. These tools are typically included with computer-aided design software for gene editing or DNA cloning. In contrast, Gelbox provides multiple dynamic representations of a gel under a wide range of conditions. Molecules can vary in size, concentration, shape, or degree of aggregation and degradation. To our knowledge, Gelbox is the only gel simulation tool that attempts to capture and explain how gels can produce flawed information, particularly due to user error, or fail completely.

To build the simulation model, we aimed to write code that would accurately convert the length of a DNA fragment to band mobility in the gel. We explored models from Van Winkle, Beheshti, and Rill¹², and the dissertation of A. Beheshti¹³, but ultimately developed our own model directly from gel-image data for a 1 kb ladder.

Gelbox has several limitations. First, Gelbox only models agarose gels and DNA fragments ranging in size from 100 to 14000 base pairs. It does not yet support RNA or protein gels. Second, the simulation model is narrow in scope compared to real agarose gels. We spent most of our effort on creating the various representations and thus didn't capture the great variation in possible parameters. Third, the simulation could benefit from further calibration. We did not systematically validate all parameters with experimental data, thus the predicted band mobility and behavior may not match real-world results. Fourth, our models do not have any physical basis: that is, we do not account for kinetic or thermodynamic properties of the DNA molecules. However, Gelbox might be used as a visualization tool for other physically based simulations if they were adapted to output our text-based file format.

In summary, we have created a dynamic, interactive gel electrophoresis simulator that relies on multiple linked representations of the system to help young or new researchers achieve more accurate and useful results when using gel electrophoresis. Gelbox may help learners establish some basic skills to interpret gel data and to troubleshoot problematic experimental conditions. Many of the limitations of Gelbox could be addressed with further development, including more experimental validation and integration with other simulation tools. We hope our work inspires others to build interactive simulations for additional aspects of scientific research so that all of these efforts can be synergistic, and so Gelbox becomes more useful and accurate with further iterations.

Acknowledgements •

We thank J Brown for assistance with the app icon and introductory video, and P Nafisi for running several gels that we used to calibrate the simulation model. We thank P Rothemund and N Case for helpful testing and feedback. This work was supported by Pew-Stewart Scholars Program for Cancer Research, the National Science Foundation Center for Cellular Construction (CCF 1317694, DBI 1548297), and the UCSF Program for Breakthrough Biomedical Research, which is partially funded by the Sandler Foundation.