I opened this topic in the Brian forum since I have tried reproducing the article in Brian2. But my question is probably not very relevant to this forum. If you have any suggestions about where to ask these questions, please let me know.
Description of problem
As a part of my bachelor’s thesis, I once tried to reproduce the mentioned paper. It turns out that, to the best of my understanding, it is not technically possible to reproduce it. There are two reasons why I think like that, but I am not even close to “being sure”. I have read the article and the supporting information repeatedly, but maybe I am missing something.
1- The structure of dendritic compartments in the model is not given explicitly. For example, how do the pyramidal neurons in layer 4’s dendrites branch? Is it like a tree? If it is, what is the structure? Is it like the figure in Izhikevich’s book “Dynamical Systems in Neuroscience” page 293, or not?
I need this information to implement the second equation on supporting information page 8. (Not allowed to embed more than one figure.)
2- What is the connectivity map between different brain regions? Is the chapter “White Matter Anatomy (After Human DTI)” in the supporting information enough to understand the structure?
Apart from the questions you’ve raised, the computational needs for the model seem to be vast:
The program simulating the model is written in C programming language with MPI and it is run on a Beowulf cluster of 60 3GHz processors with 1.5 GB of RAM each. Most of the simulations were performed with one million neurons, tens of millions of neuronal compartments, and almost half a billion synapses. It takes ≈10 min on the cluster to initialize the model, and one minute to compute one second of simulated data using a sub-millisecond time step.
It would probably need the original code to get all the details.
I have tried to reach Izhikevich by email but couldn’t get any responses. It is very unfortunate that this famous paper’s code is not available online, and nobody has published any re-production code. It would be fantastic if anyone could reach Izhikevich and get the code.
1 Like
My 2 cents on this: I wouldn’t waste any time trying to reproduce the paper – I think it is impressive from a technical point of view, but the scientific value is not great. Apart from that, I share your impression that it will be impossible to reproduce the simulations from the information given in the paper/appendix (and, as mentioned in other threads, the detailed results will most likely depend on idiosyncratic code behaviour, see Frontiers | Reproducing Polychronization: A Guide to Maximizing the Reproducibility of Spiking Network Models). The paper says “Materials and methods necessary to reproduce the results are available upon request.”, but I wouldn’t be very optimistic about it – Izhikevich left Academia quite a while ago…
Also, as @schmitts mentioned, the large scale simulation would need a computer cluster and MPI to simulate – Brian isn’t the right tool for that.
PS: I’ve moved your post to the #science category. From my point of view, posts like this are completely valid and relevant in this forum, I just wouldn’t put them into the #support category which is about concrete Brian-related questions.
2 Likes
Oh, I was unaware of the “Materials and methods necessary to reproduce the results are available upon request.” statement. Therefore, I have spent a lot of time re-reading and trying to understand which parts I am missing. So, I am giving up on trying to reproduce it.
But if someone already has these “materials and methods”, please try to put them online. Or if someone still has contact with Izhikevich, please try to ask for it.
Thank you very much for the replies and for moving this topic to the relevant category!
2 Likes
