Michael Roukes Caltech/Physics FUNCTIONAL ......Michael Roukes Caltech/Physics Thanos Siapas Caltech/Neuroscience Andreas Tolias Baylor CM/Neuroscience Gilles Laurent MPI for Brain

Michael RoukesCa l tech /Phys i c s

Thanos S i apa sCa l t e ch /Neu r o s c i e n ce

Andreas To l i a sBay l o r CM/Neu r o s c i e n ce

Gi l l e s Lauren tMPI f o r B ra i n Re sea r c h

Co-Proposers –BRAIN In i t i a t ive :

Pau l A l i v i s a tos Be r ke l e y / LBL

Miyoung Chun Kav l i F ounda t i o n

George Church Har va r d

Ra lph Greenspan UCSD/Kav l i

Ra f ae l Yus te Co l umb i a /Ka v l i

FUNCTIONAL CONNECTOMICS: TOWARD AN ACTIVITY MAP OF THE BRAIN

Michael Roukes, Caltech / 5 June 2013 © 2013 Caltech, Baylor College of Medicine, UC Berkeley Page 1 of 30

WE NEED NEW METHODS

MIchael Roukes. Caltech – © 2013 25 June 2013

overarching goal: to understand how the brain worksthe problem: “emergent” properties of brain function (that involve vastly different length scales)

What apparatus, in general terms, enables the brain to implement its remarkable performance?

The number of components (neurons) in the brain is probably about 1011 (a hundred bi l l ion) . The number of synapses, or contacts , between them is perhaps 1014 (a hundred tr i l l ion) . On the average ever y neuron receives some thousands of dist inct inputs and itsel f connects to many other neurons.

The physical layout of most of the components is not par t icular ly neat .

How might one make some headway through this impossible jungle?

Francis H Crick (1979) Think ing about the bra in . Scient ific Amer ican 241: 219–232


Evolving technology that enables complex functional imaging of the brain…

  The Human Connectome Project (e.g. Van Essen, et al .)

  Optical Reporters for Observing Brain Circuit Activity

  The Brain Activity Map Project*

5 June 2013 MIchael Roukes. Caltech – © 2013 3

APPROACHES / PARADIGMS / ANALOGIES

*Alivisatos, A.P., Chun, M., Church, G.M., Greenspan, R.J., Roukes, M.L., and Yuste, R. (2012) The Brain Activity Map Project and the Challenge of Functional Connectomics.

Neuron 74: 970-974.


f MRI-basedimaging of wholebrain activity

  The Human Connectome Project (conceptual ized representation)



analogy:computer network

“regional processors” in the brain






•  imaging of local computational activity via energy consumption









•  observation in correlations of computational activity across regions







•  observation in correlations of computational activity across regions

•  Elucidation of trans-regional connections via such correlations




thenetwork


  The Human Connectome Project

  Brain Circuit Activity via Optical Reporters



What’s going on within each of the brain’s local circuits?(it “regional processors”)

Time evolution of logic states at multiple gates


analogies

logic gate = neuronbits = neuronal “spiking”



OPTICS: FREE-SPACE FUNCTIONAL IMAGINGTHE STATE-OF-THE-ART

3-D Random Access Two-Photon ImagingWith Acousto-Optical Deflectors (AOD)

Two-Photon Functional Imaging of Visual Cortex

i.e. watching brain processing in real time!

Tolias and Saggau Groups, Baylor CM



OPTICS: FREE-SPACE FUNCTIONAL IMAGINGTHE STATE-OF-THE-ART

MAJOR PROBLEM: Calcium recording is, at present, not fast enough to faithfully record the “bits” (neuronal spiking in real time).

Two-Photon Functional Imaging of Visual Cortex

Tolias and Saggau Groups, Baylor CM

Optical recording (via calcium influx)

Simultaneous electrical recording

L. Moreaux & G. Laurent, Estimating firing rates from calcium signals in locust projection neurons in vivo, Frontiers in Neural Circuits 1, 2 (2007)

But, opticalrecordingis too slowto followfull details of neuronalspiking


Proprietary © Caltech 2011-2013 Center for Neuro/Nano Large-Scale Integration 11

MechanicalForces

ChemicalConcentrations

Electric Fields

…at each point within, there is a multiplicity of information to be gleaned



But sub-cortical brain activity cannot be accessed by free-space optics

We posit: The only near-term, general, and massively-multiplexed solution, producible en masse, is a neural-probe-based architecture

inaccessible via free-space optics

can watch here(cortex)

but not in here



nanotechnologies can enable real-time, high resolution sensing of these fields

MechanicalForces


Electric Fields

integratedneuroelectronics


MIchael Roukes. Caltech – © 2013

Toward real-time observation of highly correlated “bit trafficking” (neural coding at the individual gate level)

  The Human Connectome Project (e .g. Van Essen, et al .)

  Brain Circuit Activity via Optical Repor ter s  The Brain Activity Map Project

5 June 2013


What’s going on in the brain’s local circuits?(its “regional processors”)

Time evolution of logic states at multiple gates

analogies

logic gate = neuronbits = neural “spiking”


transmitting neuron

receivingneuron

14

individual neuronal

spikingMichael Roukes, Caltech / 5 June 2013 © 2013 Caltech, Baylor College of Medicine, UC Berkeley Page 14 of 30

15Proprietary © Caltech 2011-2013 Center for Neuro/Nano Large-Scale Integration

Roadmap: Electrophysiology

15

Questions:(a) Can we realize >50,000 channels?(b) How can we access other physical “brain fields”?(c) And, can we address (a) & (b) to permit large-

scale integration and production en masse?

>50,000 channelsmass-produced

and fully integrated100 channels1000 channels

custom

generation 1 generation 210,000 channelsmass-produced

currentstate of the art

goal

  Integrated Neuro/Nanoelectronics Roadmap:

We are here.



Next-Gen Brain-Machine Interfacing

16

integrated nanophotonics

nanoparticle-basedfunctional optical reporters+

integrated neurophotonics

in the optical domain



Neuro/Nanoparticle Reporters Under Investigation

E

- - + + + +

λem= ~573 nm λem= ~568 nm

CdSe CdS CdSe CdS

Alexandra Courtis

Mechanism of Electric Field Sensor:Quantum Confined Stark Effect

CdSe CdS

λem= >650 nm λem= ~650 nm

F

Tetrapods are Sensitive Local Force Detectors

Alivisatos Group, UCB



nanotechnologies can enable real-time, high resolution sensing of these fields:

MechanicalForces


Electric Fields

integratednanoelectronics

integratednanophotonics

functionalnanoparticles+

INTEGRATED NEUROPHOTONICS


19

>50,000 channelsmass-produced

and fully integrated100 channels1000 channels

custom

generation 1 generation 210,000 channelsmass-produced

currentstate of the art

goal

Proprietary © Caltech 2011-2013 Center for Neuro/Nano Large-Scale Integration

Roadmaps: 10Y Horizon Integrated Neuro/Nanoelectronics

19

Probe Based, LocalizedTwo-Photon Excitation

Massively-Multiplexed Optogenetic Stimulation

Next-Gen NanoparticleOptical Reporters

Two-PhotonFunctional ImagingAnd Stimulation

Probe-IntegratedMicrofluidic Deliveryand Chemisensing

 Integrated Neurophotonics


f MRI-based imagingof brain “processor” connectivity

resolution: >100K neurons/voxel

5 June 2013 MIchael Roukes. Caltech – © 201320

AN EVOLVING EXPERIMENTAL TOOLKIT

Two-photon functional imagingof brain activity

resolution: single neurontime response: seconds

Highly-multiplexed,nanoprobe-based functional imagingof brain activity

resolution: single neurontime response: ~10’s μs

nanowire arrays

recording sites

10 µm

proposed development via brain activity map project


Goal 1: Measure every action potential for every neuron in complete brain circuits

Goal 2: Manipulate the activity of every neuron in these circuits

Goal 3: Computationally analyze/model these circuits

Goal 4: (concurrent!) Develop next-gen technological platforms to permit scale-up to ever-larger hierarchical brain structures

BRAIN ACTIVITY MAP …GOALS WE ORIGINALLY ENVISAGED

MIchael Roukes. Caltech – © 2013 215 June 2013worm fly fish mouse


TECHNOLOGICAL PARALLEL:THE HUMAN GENOME PROJECT

technologyevolution


Single InvestigatorLab

Academic Center

Large-Scale Commercialization

& Production

Genome Centers


…lessons learned: HASTENING TECHNOLOGICAL EVOLUTION

phases of technological

evolution

12

43

University research:  Single-investigator development

& prototype demonstrations of “mom & pop” technologies

  One-of-a-kind prototypes,   Not amenable to near-term

automation or scale-up

University CENTERS research:

  Standardization of technologies

  Protocols become readily adaptable to automation and production en masse

Corporate spin-offs

  Production of automated systems based on “core technology” that is:

  Robust  Scalable  Mass-producible

Data Centers (“Brain Observatories”)

  Creation of massively-parallel data acquisition “centers”, which amass and synchronize the operation of large automated systems arrays; enables collection of unprecedented data libraries.

SCALE-UP “REAL” UTILITY

IP transfer



A DRAFT ROADMAP


15 years: Entire brains behaving

10 years:1 million neurons

5 years: 50,000 neurons


  Nanoscientists ; and engage their key exper t faci l i t ies and staf f

  Chemists/biochemists developing nanopar t ic le and molecular repor ter s

(for st imulat ion and recording)

  State-of-the-ar t microchip research foundries to translate “one-ofs” into

prototypes capable of scale-up and production en masse

  Industrial partners to subsequently enable mass production and system

integrat ion: enabl ing real izat ion and deployment of robust , integrated

measurement instr uments

WHO MUST WE ENGAGE TO MAKE THIS HAPPEN?


and then…

TOO

L D

EVEL

OPM

ENT


  Nanoscientists ; and engage their key exper t faci l i t ies and staf f

  Chemists/biochemists developing nanopar t ic le and molecular repor ter s

(for st imulat ion and recording)

  State-of-the-ar t microchip research foundries to translate “one-ofs” into

prototypes capable of scale-up and production en masse

  Industrial partners to subsequently enable mass production and system

integrat ion: enabl ing real izat ion and deployment of robust , integrated

measurement instr uments

WHO MUST WE ENGAGE TO MAKE THIS HAPPEN?


  Experimental neuroscientists explor ing worm, fish, mouse , rat , tur t le , …and, when wel l-val idated, pr imates and humans  Computer scientists at the forefront of massive data mining technologies  Computational neuroscientists bui lding models/analyses with next-gen complexity

and then…

TOO

L D

EVEL

OPM

ENT

NEW

SC

IEN

CE


 Devices and techniques for diagnosing brain disorder s ear l ier and more accurately

  Strategies for high-resolution patterned brain stimulation to

rebalance diseased circuits, i .e . next-gen brain-machine interfaces

  Sensit ive , miniature , and intel l igent nanosystems for engineer ing

and environmental applications

 Development of novel, biological ly-inspired, next-gen

computational devices and architectures.

AMONG THE LARGER BENEFITS…



 The BRAIN Initiative needs your support!

  Large scale: it’s too large for single-investigator grants or foundations alone. Truly big science… think satellites , telescopes, Human Genome Project.

  Long term: Horizon is too far for ear ly stage corporate or VC investment. We are talking 5, 10, 15 year horizons. Must star t now though!

  A fundamentally cross-disciplinary effort: no obvious single agency home.

 MISSION CRITICAL: The new nanotechnologies must be assembled first!

  THEN: Integrated nanosystems can provide platforms enabling entirely new ways of doing things – hastening a next generation of neurophysiological discover y

 Human genome-like project in scope – but unlike HGP we know at the outset what technologies to invest in to make the BRAIN Initiative happen!

HOW DO WE GO FORWARD FROM HERE?



NEW TOOLSNew directions in science are launched by

new tools much more often than by new concepts.

The effect of a concept-dr iven revolution is to explain old things in new ways.

The effect of a tool-driven revolution is to discover new things that have to be explained. ”

Freeman Dyson (1997) Imagined Wor lds

Har vard Univer s i ty Press , Cambr idge , MA


Neuroscience and nanoscience have separately evolved to the present, auspicious juncture – the synergistic opportunities at this moment are without precedent.


THANK YOU



Documents

Michael Roukes Caltech/Physics FUNCTIONAL ......Michael Roukes Caltech/Physics Thanos Siapas Caltech/Neuroscience Andreas Tolias Baylor CM/Neuroscience Gilles Laurent MPI for Brain