multi-level human brain modeling jerome swartz the swartz foundation rancho santa fe 9/30/06
TRANSCRIPT
Multi-levelHuman Brain Modeling
Jerome SwartzThe Swartz Foundation
Rancho Santa Fe
9/30/06
Multi-level Brain Modeling• Everyone agrees there ARE multiple levels of description• Science IS modeling• Science is intrinsically multi-level in nature (e.g. neurons
– behavior; genes – disease; atoms – molecules; etc.)• Understanding how the brain works means modeling the
dynamics of multi-level Information flow (not so easy!)• Defining the Information processed by each brain
element at each Level is essential• Dynamic brain modeling
will increasingly suffer
from Information overload:
Successful Modeling
New Measurements
New DynamicsPhenomena
Brain Research Must Be Multi-level
• Brains are active and multi-scale/multi-level• The dominant multi-level model: the computer’s physical/
logical hierarchy (viz OSI computer ‘stack’ multi-level description)
• Scientific collaboration is needed– Across spatial scales– Across time scales– Across measurement techniques– Across models
• Current field borders should not remain boundaries …Curtail Scale Chauvinism!
Level Chauvinism is Endemic…
• Dirac on discovering the positron: “the rest is chemistry”… molecular structure is an epiphenomenon!
• Systems neuroscience & neural networks: ‘the molecular level is implementational detail’… neural oscillations are epiphenomena
• Genetics/Evolutionary Psychology: genetic basis for behavior• Cognitive Psychology: largely ignores the brain itself• Almost everyone: quantum phenomena are irrelevant to biology
To progress beyond this, we must ask if there are any invariant mathematical principles underlying biological multiple level interaction
Multi-level Modeling Futures I
• To understand, both theoretically and practically, how brains support behavior and experience
• To model brain / behavior dynamics as Active requires:– Better behavioral measures and modeling
– Better brain dynamic imaging / analysis
– Better joint brain / behavior analysis
• Today’s (‘hardcore’ neurobiological) large scale computational models do not (yet) explain cognitive functions and complex behavior…. Stay tuned!
• Circuit modelers mostly work on simple *physiological phenomena* that don’t directly translate into behavioral performance
• Theorists interested in cognition predominantly use abstract mathematical models that are not constrained by neurobiology
… the next research frontiers
Multi-level Modeling Futures II
• Microcircuit models of cognitive processes (relating microscopic-to-macroscopic) to link the biology of synapses and neurons to behavior through network dynamics
• Cognitive-type circuit models detailed enough to account for neuronal data and high-level enough to reproduce behavioral events correlated to EEG and fMRI measurement and provide a unified framework
• Linear filter models are powerful for sensory processing, but cognitive-type computations involving nonlinear dynamical systems, multiple attractors, bifurcations, etc., will play an important role
Multi-level Modeling Futures III
• How do top-down ‘cognitive’ signals interact with bottom-up external stimuli? How do signals flow in a reciprocal loop between thalamocortical sensory circuits and working memory/‘decision’ circuits
• Another challenge is to expand circuit modeling to large-scale brain networks with interconnected areas/‘modules’
Multi-level Open Questions I
• Is there a corresponding (comparable?) temporal scale to our spatially-scaled Multi-level description ?
• At what time scales does Information flow between levels (how fast up & down?)?
• Are local field synchronies multi-scale?• Do local fields index shape synchronicity? • Are there any direct relationships between these
processes and nonconscious/conscious mental processing…. e.g. ‘Aha!’/‘eureka’; ‘REST’; selective attention; decision-making; problem solving; etc.
Multi-level Open Questions II
• How does Information cross spatial scales?– Up
• Spike & decision ‘ramp-to-threshold’• Stochastic resonance?• Avalanche behavior?• Within & between area synchronization avalanches?
– Down• Synaptic reshaping • Frequency nesting• Ephaptic and neuromodulator influences
Organisms
Neurons
Membrane Protein Complexes
Macromolecules
emergence boundarycondition
behavior
spikes
conformationalchanges
Information Flow in the Levels-hierarchy
Cortical hemispheres Cerebral cortex (ACC,PFC, etc.)Thalamus/sensory afferentsHippocampus-working memorySensorimotor system
Human Multi-level (“Brain Stack”) FrameworkLevel Additional Description Components Spatial Scale
Hum
an B
ehav
iora
l Lev
els
Info
rmat
ion-
The
oret
ic/S
yste
m L
evel
sP
hysi
cal/C
odin
g Le
vels
Social Neuroscience(Neuro-anthropology)
Human Interaction(Physical/Electronic)
Cognitive/Psychological(Whole Brain)
Socio-Political(Geographical/Cyber)
Neurophysiological(Anatomical
“maps”)
Network
Circuit
Neuronal
Synaptic
Molecular
Evolution-driven m:n (many:many)Global/Nation-States
Closed System Interconnect Model
Evolution/macro-plasticity km-MMm
Emotional/Rational/Innerthought
1:1 (one:one)“mirror neurons”Evolution-driver
“Network of Networks”/CNS
Communication/System sublevels
Macrodynamics
Interneuronal sublevelSynaptic/axonal/dendriticMyelination/ganglia
Neurogenetic sublevel
Physical/coding sublevel
Cortical microcircuitsThalamocortical circuits
(1k neuron) Mini-columnsNeo-cortical columns (10-100k)Synfire chains
[
1:self Conscious sublevel (presentation sublevel)
Unconscious processing
(MM: million)
dm-MMm
1 m
1cm-dm
1cm-dm
1mm-cm
1 μ -100 μ
1 Å
]
[NeuromodulatorsProteinsAmino Acids
[ ]
]
]
]]
Emotion Language Decision making (“Thin/thick slices”)Attention/awarenessSleep/awake
[ ]
[
][
[ ]
1:n (one:many)Regional/cities
Cellular microdynamic levelSpike time dependent plasticity/Learning
[
[
][[
Mic
rosc
opic
Mes
osco
pic
Mac
rosc
opic
[
[