Cellular Networks

1
Cellular Networks
Use locks and keys toghether with R and F
conjugation to build feed forward networks of
cells
2
Changing connection strength
Both connections of equal strength
Connection between cell1 and cell3 is stronger
3
Graded response to input
In liquid culture of 1/3 cellA 1/3 cellB, 1/3
cell2 expression of cell2s output is P(cellA
conjugating with cell2) Which in a well mixed
culture is proportional to the concentrations of
cellA and cell2
Pretend graph of output is here
Cell 2 produces output when it receives key 2
4
Graded response to input
In liquid culture of 1/3 cellA 1/3 cellB, 1/3
cell2 expression of cell2s output is Cell2 out
P(CellA OR CellB conjugating with cell2)
Pretend graph of output is here, higher output
than just A alone
Cell 2 produces output when it receives key 2
5
Inhibitory Signals

• Name of the protein that turns off the cell
  pili to stop receiving input but still allow
  output
• Digest/Degrade output plasmid
• Conditional cell death
• RNA based competition for key binding sites

6
What we have

• Addressable communication
• Hierarchical network architecture
• Adjustable connection strengths
• Graded aggregate response to input
• Inhibitory signals

All the components required for a feed forward
neural network
7
Back Propagation Neural Network
Input signals propagate forward increasing
activity, both positive and negative
Error signals propagate proportionally backwards
returning activity to 0
8
General Node Design
Receive input from many inputs, send output to
many outputs, relay error from many outputs to
many inputs
9
Bacterial Neural Networks

• Massively Parallel
• Probabilistic
• Asynchronous
• Continuous time
• Can be tied into other pathways in cell or
  environmental conditions
• Highly adaptive, can grow additional nodes
• Complex behavior from simple, uniform node design
  with different lock/keys