Network analysis of mesoscale mouse brain structural connectome yields modular structure that aligns with anatomical regions and sensory pathways

Abstract

The Allen mesoscale mouse brain structural connectome is analysed using standard network methods combined with 3D visualizations. The full region-to-region connectivity data is used, with a focus on the strongest (top 40%) structural links. The spatial embedding of links and time evolution of signalling is incorporated, with twostep links included. Modular decomposition using the Infomap method produces 8 network modules that correspond approximately to major brain anatomical regions and system functions. These modules align with the anterior and posterior primary sensory systems and association areas. 3D visualization of network links is facilitated by using a set of simplified schematic coordinates that reduces visual complexity. Selection of key nodes and links, with a focus on primary sensory pathways and cortical association areas, together reveal structural features of the mouse structural connectome consistent with biological functions in the sensory-motor systems, and selective roles of the anterior and posterior cortical association areas of the mouse brain. The fabric of weaker links generally are longer range with some having brainwide reach. Cortical gradients are evident along sensory pathways within the structural network. Author's SummaryNetwork models incorporating spatial embedding and signalling delays are used to investigate the mouse structural connectome. Here computational method work like experimental probes to uncover biologically relevant features. I use the Infomap method, which follows random walks on the network, to decompose the directed, weighted network into 8 modules that align with classical brain anatomical regions and system functions. Primary sensory pathways and cortical association areas are separated into individual modules. Strong, short range links form the sensory-motor paths while weaker links spread brain-wide, possibly coordinating many regions. modern experimental probes (e.g. see [8]). The challenge is to find appropriate network models that can realistically reproduced/predict brain structural and functional connectivity data.Here the Allen Institute for Brain Science (AIBS) dataset of a mesoscale structural connectome for the mouse brain [9] is studied using a combination of data analytical, network and visualization methods. While the data is brain wide it is at the meoscale (resolution of 0.1 mm voxels and 0.35 mm coronal plane imaging), in which links involving many neurons are probed. Connectivity data at finer scales, at the level of individual neurons and synapses, are available for smaller samples, such as with mouse retina [10,11]. Higher resolution data also is emerging at the level of cortical layers [14]. These studies shed light on the correlation between structural and functional connections between nodes in the mouse brain. The nodes are located at known 3D coordinates from the Allen Atlas [13]. Typically the sources and targets of the probed connections are classical anatomical regions that, given sampling protocols, vary significantly in size and fu...

Network analysis of marmoset cortical connections reveals pFC and sensory clusters

“…1 suggest that some in links of low-medium weight are missing. A similar plot for mouse brain [Pailthorpe 2019] shows both distributions being quite similar - again suggesting some missing in-weight data here. This is consistent with sampling only 55 source nodes, being the 55, of the 116, anatomical areas used as tracer injection sites.…”

“…al. 2014, Pailthorpe 2019] and EM based results for worm [Varshnay 2011] and mouse retina [Helmstaedter et. al.…”

“…al. 2014, Pailthorpe 2019], but in contrast to the EM measurements of synaptic contacts or areas in mouse retina [Helmstaedter et. al.…”

“…al. 2014, Pailthorpe 2019]. However the original weighted in-degree adopts only values of 0 or 1 since fractional measures sum to 1 for each node, which follows from the definition of FLNe.…”

Simulated dynamical transitions in a heterogeneous marmoset pFC cluster

“…Network analysis (Pailthorpe, 2024 ) of the marmoset cortical connectivity data, measured by retrograde tracers (Majka et al, 2016 , 2020 ), identified a significant 3D cluster in and around the pre-frontal cortex (pFC), which is examined here. The network analysis methods followed those used previously for the mouse retina (Pailthorpe, 2016 ) and the mouse brain (Pailthorpe, 2019 ). Previous studies of brain networks (Sporns and Kotter, 2004 ) generally focused on small motifs comprising 2, 3, or at most 4 nodes due to the combinatorial explosion in the number of possible configurations of larger clusters.…”