The Virtualized Macaque Brain
By Kelly Shen
In my last blog post, I wrote about how we were in the midst of building a large-scale connectome of the macaque brain as an extension of TheVirtualBrain. Well, at long last, it’s ready and available for all of your connectome-based modeling efforts!
Constructing a connectome, as many of you know, is nontrivial. In the macaque, several large-scale connectomes already exist from tracer studies and they offer specificity not available in connectomes reconstructed using diffusion-weighted imaging (DWI) tractography. The problem, however, is that in connectome-based modeling, white matter fiber tract capacities (i.e., edge weights) and tract lengths are critical for scaling the spatiotemporal interactions of the network. Neither of these are well estimated by the existing large-scale macaque connectomes. So, we took advantage of an existing whole-cortex tracer connectome derived using the CoCoMac database and “enhanced” it using DWI tractography in macaques to estimate the weights and tract lengths.
The Virtualized Macaque Brain (essentially a lot of hard work reduced into just two matrices) is used to specify connectivity of coupled neural mass models in TheVirtualBrain (TVB).
The result was a fully-weighted and directed whole-cortex connectome. I want to specifically mention the directedness of the connectome here because it’s something we don’t often talk about in human connectome construction / modeling, owing mostly to the fact that DWI tractography doesn’t allow us the ability to infer direction of a connection. However, from our work in macaques, we know that connection asymmetry has quantifiable consequences on functional interactions [1,2] and that these connections make up 10-15% [2,3] of the connectome. Even when bidirectionality exists, the weights can be heavily asymmetric. So what I’m saying is that if we can, let’s not overlook the directionality of connections and the role they play in network dynamics.
You can find the details on how we built the connectome in our latest preprint here and The Virtualized Macaque dataset is available for download here. In the paper, we show some example simulations of macaque resting-state fMRI data but you could also simulate local field potentials and magneto-/encephalography using TVB. We include, as part of the package, the macaque cortical surface so you could do surface-based simulations too, if you’re so inclined. For us, this connectome, along with The Virtualized Mouse Brain, represents an important step towards bridging spatial and temporal scales using a theoretical framework that can combine empirical data from multiple scales of investigation and across animal models. But what I’m most excited for is to see all the creative and important things that you all do with this connectome. So, please keep me updated via the Twitterverse: @shen4brains
