Line 137: | Line 137: | ||
<a href="https://github.com/Manchester-iGem-2016/UoMiGem2016">6 Made a Ensemble modelling code series for other teams to learn from | <a href="https://github.com/Manchester-iGem-2016/UoMiGem2016">6 Made a Ensemble modelling code series for other teams to learn from | ||
</a> | </a> | ||
− | |||
− | |||
− | |||
− | |||
− | |||
</p> | </p> | ||
Revision as of 02:12, 19 October 2016
Welcome to our modelling section. We have used a novel ensemble modeling approach, to better aid the synergy between wetlab and dry lab teams. On this page you will find answers to the questions; What is ensemble modelling? What did we model? what did our model achieve?
For navigating the Wiki you need to know that the sections on results and human practice/lab integration can be accessed in the menu bar.
Part of what we hope to achieve with this ensemble methodology is a blueprint for other iGEM teams. As such each step in creating and using our model is laid out in the diagram you will see by clicking what is ensemble modelling, alternatively click here. Clicking on a specific step will take you to a page explaining the theory. Going deeper you can access a discussion of how to do this in practice. Click below to access our code, use as you wish.
Link to github homepage
Our main modelling achievements this summer
1 Introduced ensemble modelling blueprint to iGEM 2 Using ensemble modelling to improve our knowledge about our network topology 3 Human practices Cost analysis of our patch (police link) 4 Benefits From Parameter Relationship Analysis 5 Learnt from experiment and integrated into model 6 Made a Ensemble modelling code series for other teams to learn from
What were we modelling?
We focused on modelling the Cell-free Mechanism. The short version is the AlcoPatch relies on alcohol, alcohol oxidase (AOx), horseradish peroxidase (HRP) and 2,2'-azino-bis(3-ethylbenzothiazoline-6-sulphonic acid) (ABTS) which are mixed together producing oxidised ABTS which is colourful. The ensemble modelling process will work for any model however and this was chosen as a simpler example to demonstrate the technique, focusing on the process rather than the system itself.
Due to time constraints the modelling was based on an analogous system of glucose and glucose oxidase (GOx) rather than alcohol and AOx, this was chosen as the reaction network is the same and it links in with the suggestions given at the Microbiology Society Conference in the human practices. Namely the design need not be limited for alcohol but could be used by diabetics, etc if used to detect other molecules in sweat. While the equivalent analysis was done for glucose rerunning the analysis for alcohol would only require the change of some constants, so the analysis acts as a sufficient proof of concept and still shows the integration of human practices.
A schematic diagram of the final scheme is given below. For more information about the steps click on the blue enzyme boxes.
![Reaction Network Diagram used in the modelling](https://static.igem.org/mediawiki/2016/0/04/T--Manchester--ModellingNetworkDiagram.png)
Alternatively you can click on the enzyme name below:
What is Ensemble Modelling?
Instead of running our model with a set of specific parameters (for example rate constants.), we run our model multiple times using different sets of parameter values and analysed the predictions as an ensemble. We collected all the possible parameter values from published literature and took into account the uncertainties that are associated with them. These parameters where described by probability density functions. This has created probabilistic outputs allowing us to make rigorous conclusions about our reaction mechanism. Ensemble modelling has yet to ‘breakthrough’ into the world of biological modelling and systems biology in general. Essentially, we analyse multiple predictions from our model as a composite, which is different from traditional predictive modelling methods. Instead of running a single prediction from a fixed single set of kinetic parameter values, we run our AlcoPatch model a number of times from different sets of kinetic values by sampling from the probability distributions generated from the previous step. We then analyse the entire set of predictions as an ensemble to draw probabilistic conclusions about the system. To explore the theory of this process please click the boxes on the diagram below.
![Overview flowchart of ensemble modelling](https://static.igem.org/mediawiki/2016/a/a9/T--Manchester--ModelFlowchart.jpg)
Alternatively you can click on the step name below:
Collecting and Processing Data Generating Probability Density Functions Simulate the System Analyse the Results Story of the Model
What did our model achieve?
We had 3 main aims for our model: to introduce Ensemble Modelling to iGEM, to improve our understanding of our system, and hence the model, and to answer key questions that arouse during the Human Practices. Below you can find the results of the latter two aims:
Network Mechanism Analysis | Parameter Relationship Analysis | Cost Analysis |
---|---|---|
Comparing model predictions with experimental data for different potential circuit topologies | Assessing the interlinking nature of specific parameter pairings on the outcomes of the system | Predicting the costs for a range of different system specifications by varying the amount of enzymes based on experimental data |
We found great inspiration from our human practices and guidance working both ways with the experiments. Click here to see a summary.