Cytoscape - Working with thousands of networks

Question

[David]

Hello everyone,

I am working with Cytoscape to study the 3D conformations of oligomers. Basically, we perform molecular dynamics simulations and save one snapshot (one conformation, one "3D structure") of the molecule each nanosecond. Then, we can isolate one snapshot of our molecule and create a network where all heavy atoms are nodes. An edge is considered when the distance between two nodes is below a threshold. We can then create a .graph file which contains the list of connected nodes, by pairs. This .graph file is then analyzed with Cytoscape as a network, to see how are connected the different atoms of the molecule.

The thing is, in our molecular dynamics simulations, we have thousands of nanoseconds and thus thousands of snapshots. I would like to know if it would be possible to import all the .graph files at once with the same "import parameters" instead of doing it for each individual file (import, selecting "source nodes", "target nodes"). All the .graph files contain 3 columns: the first one contains the "source nodes" and the second one contains the "target nodes" (the third one can be ignored). Also, the Advanced Options would be the same for all files: space as a delimiter, dot as decimal separator, "Use first line as column names" unchecked. I guess that a simple script could do the job but I have very few knowledge in scripting/programming. If you have any idea of how to automatize it, please let me know, as it would be very tedious to do it manually for thousands of files.

The idea then would be to import all those networks and perform a merging of them all (with the Merge built-in tool). Concerning merging, would you have any additional information about how it is done in more details? The manual does not contain much information about how it works exactly. Is merging indeed a good way to get an "average network"?

Thank you very much in advance for your time and do not hesitate to tell me if there is anything unclear in my question, Best regards,

David.

Answer 1

[Matthias Zepper]

Admittedly, it is more than a decade ago that I last worked with 3-dimensional molecule data, but I would suggest to rethink your approach. Cytoscape is a network visualization and analysis tool e.g. for pathways, which can possibly be used in the described manner, but I would refrain from doing so. You are throwing so much information away by coercing it into a binary edge/no edge format, e.g. if certain events are mutually exclusive or happen always simultaneously.

If I understand you correctly, you are trying to plot/visualize the most frequent conformation of your molecule and possibly highlight atoms that are frequently below a specific threshold distance? Why don't you use e.g. UCSF Chimera for that? Have for example a look at the manual about Structure averaging or superimposing structures, the Ensemble Cluster and Match functionalities. For visualization of the trajectories, you can e.g. use TileClusters.

Maybe there is in the meantime better and more modern software available than UCSF Chimera, which I used back in the day, but it is definitively a very powerful tool suite.

[David]

Thank you for your quick and helpful answer!

The idea is indeed to highlight atoms or groups of atoms that are strongly connected. In fact, we are studying molecules bearing functional moieties that need to be close to each other to be active. So the goal is not really to have a precise idea of the whole conformation of the molecule but rather to get insights on the connectivity of specific parts of the molecule along the trajectory. Using Cytoscape and the network representation is interesting as it adds a "coarse-grained" view of the system, in addition to the all-atom model (which is also analyzed).

Thank you for the suggestion of using Chimera, I will definitely take a look at it. In fact, using cpptraj (the analysis tool of Amber), it is already possible to do some of the things you mentionned (the clustering for example could be very useful, but as different algorithms and parameters can be choosen, it requires some work to get it right).