Feasibility Study
RQ2-Feature Models
RQ2: What are the characteristics of feature models built from a DNA repository?
Our complete workspace for all models in FeatureIDE can be downloaded here. We also include jpg images for each feature model. To view the complete feature model click on the images below.
Setup
FeatureIDE is an eclipse plugin for feature modeling. Installation instructions for FeatureIDE can be found here. It is also an available software package and is searchable under Help --> Install New Software. In case of problems please refer to the FeatureIDE documentation. Please note as as June 2019, FeatureIDE is compatible only with Eclipse versions through 4.9.
In this work we used:
Eclipse IDE for C/C++ (Developers Version: 2018-09 (4.9.0) Build id: 20180917-1800)
FeatureIDE v3.x (http://featureide.cs.ovgu.de/update/v3/)
CDT (http://download.eclipse.org/tools/cdt/releases/9.5)
To use our workspace, download and unzip the directory linked above. When you open Eclipse and are prompted to select a workspace, browse for the directory and select it (the entire directory named featureide-workspace). If for some reason you are not prompted for a workspace go to File --> Switch Workspace --> Other. When you Launch all models should be loaded.
Cell to Cell Signalling (Figures 8 & 9)
To create the cell-to-cell signalling model we manually analyzed all basic parts within the Cell-cell signaling and quorum sensing function category on the Registry of Standard Parts. Two categories of required parts were found in separate webpages: Ribosome Binding Sites and Terminators. The developed feature model can be seen below (click on the image to view the full model).
Due to the scale of this model we had to use FAMA Version 1.1.2 [1] to calculate the number of products. The input files for each model can be found in our artifacts can be found here (the FAMA tool downloads can be found here). Due to the scale we split the model into the three top level features (sender, receiver, reporter) and ran them separately. The total number of products was found by multiplying them. The cell-to-cell signalling model was run on an HPC cluster with 35GB and java was given 32GB (java -jar ./lib/FaMaSDK-1.1.1.jar -xmx 32). The sender and receiver each have 11,448,000 products and the reporter has 5,724,000 products. Multiplying 11,448,000 * 11,448,000 * 5,724,000 = 7.50169 × 1020 products for the whole model.
Kill Switch (Figure 10)
We manually reviewed the wiki pages for the 110 Gold Medal iGEM teams from the 2017 competition. We flagged those teams who indicated that they contemplated or actually implemented a kill switch in their projects. 15 of them mentioned a kill switch, but only 12 were picked as having actually attempted an implementation. All teams can be found in this spreadsheet. We then reviewed their descriptions of their kill switches to determine which BioBricks they used for implementation, some of which were novel parts submitted for addition to the registry. Based on their descriptions of the triggering conditions and the specific mechanisms causing cellular death, we manually built the feature model. We calculated the number of products using FeatureIDE (882 products).
Viral Vector
The set of viral vector parts in the BioBrick repository is a set of 103 parts that can be used to create a version of the Adeno-associated virus (AAV2). All but one of these parts (which we eliminate from our set of parts) was added to the repository by the 2010 iGEM team from Freiburg (Freiburg Bioware 2010). The Freiburg team created a “Virus Construction Kit” to allow other users to create an AAV2 virus. To build the feature model, we use the list of parts in the BioBrick repository catalog page, and domain knowledge from the Freiburg iGEM team including their “Virus Construction Kit Manual.” In review of the Freiburg team we identified two main components of a viral vector system, the gene of interest (GOI) vector and the capsid vector. The GOI vector represents the gene of interest that will be transported into the host. The capsid vector represents the container that the gene of interest will be transported in. We consider these two separate feature models. We also employ domain knowledge from a domain expert.
Viral Vector - GOI (Figure 11)
Viral Vector - Capsid (Figure 12)
