Difference between revisions of "Team:Vanderbilt/Software/Development"

The wetware team devised multiple methods to identify and remove what they called "mutation hotspots," or specific sequences of DNA known to be susceptible to certain types of randomized damage. Through a set of experiments, the team demonstrated that these methods of removing hotspots reduced the frequency of mutation in the parts produced. We denoted the process of removing these mutation hotspots to produce a more stable part "mutation optimization." Jarrod Shilts wrote a Python script to automatically perform this operation, which was used in the wetware team's experimentation. The software portion of this project entailed porting the optimization script from Python to JavaScript, and producing a library, console application, and publicly accessible web application.

Like the wetware team, the software portion of this project is organized into both sequence and circuit optimization methods. The sequence optimization section presents a simple interface with smart defaults. An example is shown below.

The user pastes or drag-and-drops the original sequence into the box on the left, and the optimized sequence appears on the right. The optimized sequence is heavily annotated as shown with all of the changes the optimizer made while it worked. A variety of parameters are available to the user, detailed in the "Documentation" link at the top.

The application takes a "plug-and-play" approach, where knowledge of the underlying principles isn't required to effectively operate it. Because of the genericity of the mutation hotspots identified by the software, it operates well on a large variety of inputs without any user interaction. The capabilities of the software are more thoroughly discussed in the wetware team's documentation.

The circuit optimization kit takes a similar approach, applied to higher-level parts instead of individual sequences. The user inputs a list of part numbers, and the software will suggest replacements for parts which conflict to cause a higher probability of mutation. More documentation about the capabilities of the circuit optimization interface is available on the wetware team's page. An example of the circuit optimization interface is shown below.

While the wetware team demonstrated that optimizing a sequence reduced its frequency of mutation, the question was left outstanding whether many parts actually could be optimized. To determine this, we mined the Parts Registry and ran our optimization algorithm on a large number of parts. Our results are available on the Stats page. As discussed above, this required creating a wrapper over the Parts Registry, which we released as a separate project to be used by other teams.

We chose to translate the code to JavaScript so that we could create an interactive web application for scientists, while keeping an extensible command-line application and software library open for other developers, using the same codebase. JavaScript is also much faster than Python in many sequence operations, which reduces the significant time required to optimize a part. To support this effort, we created and released a generic library of optimization tools for JavaScript applications (see DISCOTANGO).

In addition to creating user-facing applications, we also collected statistics on the efficacy of the mutation optimization scheme on publicly available biobricks. This required mining the registry for part information, which was a challenge in itself. We wrote a wrapper to query the Parts Registry for part information and used it to collect bulk statistics on the Parts Registry's content (see biobrick-api-wrapper).