Design for a Genetic System Capable of Hebbian Learning Chrisantha Fernando Systems Biology Centre Birmingham University January 2006 Chrisantha Fernando.

Design for a Genetic System Capable of Hebbian LearningDesign for a Genetic System

Capable of Hebbian LearningChrisantha Fernando

Systems Biology Centre

Birmingham University

January 2006

Chrisantha Fernando

Systems Biology Centre

Birmingham University

January 2006

Hebbian LearningHebbian Learning

Donald Hebb 1949. “Let us assume that the persistence or repetition of a reverberatory activity (or "trace") tends to induce lasting cellular changes that add to its stability. When an axon of cell A is near enough to excite a cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased.”

Long term potentiation in neurons subsequently demonstrated, along with LTD. Hebbian Learning can implement

Associative Learning, e.g. LTP in Hippocampus. Auto-associative Memory, e.g. Cerebellar motor memories. Self-Organized Map Formation, e.g. ocular dominance columns.

Donald Hebb 1949. “Let us assume that the persistence or repetition of a reverberatory activity (or "trace") tends to induce lasting cellular changes that add to its stability. When an axon of cell A is near enough to excite a cell B and repeatedly or persistently takes part in firing it, some growth process or metabolic change takes place in one or both cells such that A's efficiency, as one of the cells firing B, is increased.”

Long term potentiation in neurons subsequently demonstrated, along with LTD. Hebbian Learning can implement

Associative Learning, e.g. LTP in Hippocampus. Auto-associative Memory, e.g. Cerebellar motor memories. Self-Organized Map Formation, e.g. ocular dominance columns.

QuickTime™ and aTIFF (LZW) decompressor

are needed to see this picture.

Examples of Hebbian LearningExamples of Hebbian Learning

Associative learning: Classical ConditioningUCS = CollisionCS = Proximity

Hebbian Learning in RabbitsHebbian Learning in Rabbits



Self-Organized MapsSelf-Organized Maps





http://web.bryant.edu/~bblais/pdf/job_talk/sld033.htm

How is Hebbian Learning Implemented in Brains?

How is Hebbian Learning Implemented in Brains?

QuickTime™ and aTIFF (Uncompressed) decompressor


Pre

Post

+/-

What about Learning in Single Cells?

What about Learning in Single Cells?

No conclusive evidence that unicellular organisms can do associative learning on their own.

Bacteria can habituate and be sensitized but they have not been shown to undertake associative learning.

No conclusive evidence that unicellular organisms can do associative learning on their own.

Bacteria can habituate and be sensitized but they have not been shown to undertake associative learning.

Vibration

Shock

Slower Replicating Eukaryotes in Complex Environments Might

Benefit from Lifetime Learning.

Slower Replicating Eukaryotes in Complex Environments Might

Benefit from Lifetime Learning.

Fine tuning of obstacle avoidance/motion. Learn that chemical A (or a spatio-temporal

pattern of A) is associated with food or harm rather than chemical B or C or D…

Achieve robust development in the face of lifetime noise due to external perturbations and internal (mutational) perturbations.

Fine tuning of obstacle avoidance/motion. Learn that chemical A (or a spatio-temporal

pattern of A) is associated with food or harm rather than chemical B or C or D…

Achieve robust development in the face of lifetime noise due to external perturbations and internal (mutational) perturbations.

Intra-cellular Molecular Networks

Intra-cellular Molecular Networks

Dennis Bray (Nature 2003). Neural networks have a remarkable ability to learn different patterns of inputs by changing the strengths of their connections (10). They are widely used in a variety of tasks of machine recognition. From the standpoint of a living cell, the closest approximation to a neural network is probably found in the pathways of intracellular signals (11). Multiple receptors on the outside of a cell receive sets of stimuli from the environment and relay these through cascades of coupled molecular events to one or a number of target molecules (associated with DNA, for example, or the cytoskeleton). Because of the directed and highly interconnected nature of these reactions, the ensemble as a whole should perform many of the functions commonly seen in neural networks. Thus, in aggregate, the signaling pathways of a cell are capable of recognizing sets of inputs and responding appropriately, with their connection "strengths" having been selected during evolution.

Eric Winfree and Hopfield view gene transcription networks as CTRNNs, pto.

Dennis Bray (Nature 2003). Neural networks have a remarkable ability to learn different patterns of inputs by changing the strengths of their connections (10). They are widely used in a variety of tasks of machine recognition. From the standpoint of a living cell, the closest approximation to a neural network is probably found in the pathways of intracellular signals (11). Multiple receptors on the outside of a cell receive sets of stimuli from the environment and relay these through cascades of coupled molecular events to one or a number of target molecules (associated with DNA, for example, or the cytoskeleton). Because of the directed and highly interconnected nature of these reactions, the ensemble as a whole should perform many of the functions commonly seen in neural networks. Thus, in aggregate, the signaling pathways of a cell are capable of recognizing sets of inputs and responding appropriately, with their connection "strengths" having been selected during evolution.

Eric Winfree and Hopfield view gene transcription networks as CTRNNs, pto.

http://www.sciencemag.org/cgi/content/full/sci;301/5641/1864#REF10








Incomplete promotor region

Sigmoidal activation function is sequence programmable buthard wired.

The New RNA World The New RNA World

John Mattick, Sean Eddy etc… are investigating RNA networks in Eukaryotes.

98% of RNAs in human cells is Intron or non-coding.

10x more RNA on Chromosome 21 and 22 is non-coding than exons.

Many types of novel RNA have been discovered capable of regulation of gene expression.

It may eventually explain the C-value paradox.

John Mattick, Sean Eddy etc… are investigating RNA networks in Eukaryotes.

98% of RNAs in human cells is Intron or non-coding.

10x more RNA on Chromosome 21 and 22 is non-coding than exons.

Many types of novel RNA have been discovered capable of regulation of gene expression.

It may eventually explain the C-value paradox.



1

What if RNA networks are really like little neural networks?

What if RNA networks are really like little neural networks?

Current neural network metaphors for cellular networks have not included plasticity.

How difficult would it be to design (or evolve) a network capable of Hebbian learning?

What sorts of tasks could an intra-cellular Hebbian learning mechanism solve?

Current neural network metaphors for cellular networks have not included plasticity.

How difficult would it be to design (or evolve) a network capable of Hebbian learning?

What sorts of tasks could an intra-cellular Hebbian learning mechanism solve?

u1 + W*1

u2 + W*2 u2W*2

u1W*1

RNA

W*2 W

W*1 W

RNA + u1 RNAu1

RNA + u2 RNAu2

Promotor Gene

k1

k2

10

10

0.005

k1

k2



Time

[Conc][RNA]

[TF1]

[TF2]

[RNATF1][RNATF2]

Stimulate with unit TF2

Stimulate with unit TF1

Experiment 1: Sensitization

Time

[Conc]

[W1*]

[W2*]

‘Weight’ Change in Experiment 1



Associative LearningAssociative Learning

UCS (e.g. glocuse) stimulates TF1, which binds strongly to the promotor and produces RNA (innately).

CS (e.g. potassium permanganate, or NO or something else) stimulates TF2 but these TFs bind only very weakly to the promotor.

Paired exposure to UCS and CS result in strengthening of the TF2 binding to promotor, and a response to the ‘smell’ associated with glucose.

UCS (e.g. glocuse) stimulates TF1, which binds strongly to the promotor and produces RNA (innately).

CS (e.g. potassium permanganate, or NO or something else) stimulates TF2 but these TFs bind only very weakly to the promotor.

Paired exposure to UCS and CS result in strengthening of the TF2 binding to promotor, and a response to the ‘smell’ associated with glucose.



Self-Organized MapsSelf-Organized Maps

Two TFs act on two genes Now add lateral inhibition Gene ‘receptive field’ changes Gene A represents TF1 level

Gene B represents TF2 level

Two TFs act on two genes Now add lateral inhibition Gene ‘receptive field’ changes Gene A represents TF1 level

Gene B represents TF2 level



Supervised Learning: Training a gene perceptron.

Supervised Learning: Training a gene perceptron.





ApplicationsApplications

Engineering cells as biosensors in complex environments. Put cell in, then measure W* levels to obtain the recorded weights.

To train a cell to produce a protein under required conditions, without having to hard-wire the promotor-TF interaction perfectly.

Engineering cells as biosensors in complex environments. Put cell in, then measure W* levels to obtain the recorded weights.

To train a cell to produce a protein under required conditions, without having to hard-wire the promotor-TF interaction perfectly.

Thanks to Thanks to

Dov Stekel Jon Rowe Bruce Shapiro Sally Milwidsky

Dov Stekel Jon Rowe Bruce Shapiro Sally Milwidsky

Design for a Genetic System Capable of Hebbian Learning Chrisantha Fernando Systems Biology Centre Birmingham University January 2006 Chrisantha Fernando.

Documents

associative learning

lifetime learning

proximity slide

hebbian learning donald

vibration shock slide

pre post slide

chemical b

cell b