Modules, Hubs and Communication Dynamics in Brain Networks: Olaf Sporns
<p>Jul 02, 2016 03:41 PM this is at USC one year ago...</p>
<p>
social networks (transport)
finance
cellular metabolic nets... proteomics...</p>
<p>
network science....</p>
<p>
get away from static concept...</p>
<p>
c elegans, drosophila, rodents, macaque;</p>
<p>
brain regions, cortical regions</p>
<p>
how to do parcellations;</p>
<p>
how strong is the connection and what direction does it go...</p>
<p>
structural connectome in matrix...</p>
<p>
versus... time series of neuronal activity... bivariate estimate ... cross correl.</p>
<p>
functional networks (and how to relate to structural)</p>
<p>
nodes and degrees... the degree is important (# of connex)</p>
<p>
influence... centrality; </p>
<p>
only 1 in 10 million in the brain (connec of neurons)</p>
<p>
communities of nodes....</p>
<p>
a hub node btwn two communities...
privileged... but vulnerable...</p>
<p>
func... statist; dynamic;
poss infinite configurations;</p>
<p>
temporal persistence;</p>
<p>
functional connectome is what the connectome DOES.</p>
<p>
*</p>
<p>
cross species...</p>
<p>
C elegans... mid 1980s... by hand.. lots of EM...</p>
<p>
some neurons have high connex... high centrality...</p>
<p>
rich club organiz.</p>
<p>
drosophila...</p>
<p>
LPU local proc unit...</p>
<p>
mouse connectome... Wash U. his work...</p>
<p>
dorsal vs ventral...</p>
<p>
high clustering in rat... small world...</p>
<p>
rich club... highly connected nodes (and they are connected to each other)</p>
<p>
van Essen Felleman... an adjacency matrix... in macaque... henry kennedy...</p>
<p>
</p>
<p>
lots of common properties....</p>
<p>
high clustering, short path length...</p>
<p>
hubs... log normal profile of connectome density...</p>
<p>
what are you connected to... who are you talking to? that's impt...</p>
<p>
*
perhaps there's a need for a rich club...
an antomic substrate where info can come together...</p>
<p>
generative models...
wiring rules... in computer program to generate these...</p>
<p>
what are the rules to gen. this network from scratch.
2 nodes that are close by are connected (an example of a hypothet rule)</p>
<p>
minimizing material (since Cajal)... eg wiring minimization</p>
<p>
but some value added when that is violated...</p>
<p>
generative models do a good job...</p>
<p>
* Human connectome *
2008 with Patric Hagmann</p>
<p>
with Antonio... precuneus, poster cingulate, etc. medial cortex... rich club...</p>
<p>
sturc core...</p>
<p>
v highly interconnected... and highly connected other hubs...</p>
<p>
posterior medial cortex... agreement...</p>
<p>
<b>participation coefficien</b>t is best way to characterize... better than degree.</p>
<p>
and implicated in pathology (are hubs)</p>
<p>
30 brain disorders... autism, schizo, alzheimers... etc.</p>
<p>
where is impairment?</p>
<p>
in highly connected parts of brain eg in MTL (and others along midline)
impaired hub regions... that's bad...</p>
<p>
the RICH CLUB he shows
netowrks</p>
<p>
densely interconnected... a drive to maintain these connections...</p>
<p>
highly influential people... given one another low rates!!! he jokes.</p>
<p>
(but not true for protein networks!)</p>
<p>
more connex among high degree nodes...</p>
<p>
post cingulate, precuneus,</p>
<p>
medial frontal cortex,</p>
<p>
insula in both hemisphers and MTL.</p>
<p>
touching rich club one node 90% of time...
sending a msg from any A to any B...</p>
<p>
shortest path....</p>
<p>
and 2 RC nodes 2/3s of time despite</p>
<p>
RC only taking up 10% of brain...</p>
<p>
soaks up traffic a lot...</p>
<p>
like hiways 40K miles...</p>
<p>
25% of cars on hiways...</p>
<p>
*</p>
<p>
RC and RSN = resting state networks...</p>
<p>
RSNs have anatomic basis...</p>
<p>
nonstationary functions in RSN...
eg Grecius.</p>
<p>
in macaque also...</p>
<p>
is the RC just one of them eg for DMN...</p>
<p>
Marty van den Heuvel.</p>
<p>
2 difft modalities: struc vs func...</p>
<p>
RC... three DMN, salience, and FPN the big three...
multimodal.. versus unimodal networks (vision, audition)</p>
<p>
heavily anchored in cognitive networks in the three above...</p>
<p>
how are these generated (the three above)...</p>
<p>
functional activ is fluctuating a lot from second to second... but shaped by
anatomy...</p>
<p>
RSN intrins. connected...</p>
<p>
dramatic effects from callosotomy eg.</p>
<p>
hard to rpedict inverse (going from func to struc)</p>
<p>
want to gener time series from struc models...</p>
<p>
use CoCoMac to gener func series.</p>
<p>
use biophys eqns...</p>
<p>
get synth fMRI doing MRI without a scanner...</p>
<p>
throwaway entire brain regions... much cheaper... he jokes...</p>
<p>
anesth monkeys in Japan... two...</p>
<p>
just 3 eqn per brain region...</p>
<p>
could do spiking neurons and do more accurate...</p>
<p>
but how abt... against the funding stream... he jokes "I have tenure" who cares...</p>
<p>
say brain is a commun network...</p>
<p>
topo... enables efficient communic... short paths.</p>
<p>
PNAS 111 823 in 2014... Gofi...</p>
<p>
say there are off ramps... into Times Square...</p>
<p>
you need more info ... Shannon entropy...</p>
<p>
can predict func connectivity...</p>
<p>
4 measures... strength of connection, length of path, etc.</p>
<p>
communic model... makes a great prediction...</p>
<p>
a dynamic model...</p>
<p>
convergence zones in macaque...
HC... is prominent...</p>
<p>
in vol one of neurl computation by Antonio.Damasio...</p>
<p>
Summary and conclusion..</p>
<p>
Hub nodes interconnec among each other...</p>
<p>
</p>
<p>
</p>
<p>
</p>
<p>
</p>