User:J. C. Martinez-Garcia/Notebook/HMS Activities/2008/11/05
|The network|| Main project page|
Previous entry Next entry
I worked all the day to built the artificial neural network which corresponds to the computer-based simulator of the phenomena described by the experimental data obtained when manipulating the AKT-signaling pathway. It was easy to find the right network, just using the Neural Networks Toolbox by Matlab. The chosen network reproduced in a very nice way the measured data when fed by that Zstk concentration as well as the logical variable corresponding to the pre-treated/treated case and the inhibition of mTOR raptor. I wrote a latex file which summarizes all the results. The file as well as the Matlab sesion are included in this ZIP file (simultor.tex is the LaTeX file, and ThisNetwork.mat corresponds to the network data and network_AKT.mat corresponds to the Matlab Workspace): Media:network_AKT.zip.
What to do with the network
When thinking in the utility of a neural network in the kind of research concerned by the manipulation of the AKT signaling pathway. It could be argued that it can be used as a memory of the experimental results, as a way to stock the data, which is in fact the main purpose of modeling. However, neural network are as modeling tools black boxes and because of that is difficult to have an insight about the biological dynamics underlying the observed empirical behavior. I think that this tool can be useful as a way to develop simulators in a fast way, and that can be important when developing more complex models (i.e. ordinary differential equations models). The simulator cab be used to test the complex model, which is to say to verify if the new models reproduce the data displayed by the network.