README.md

Rationale of these experiments

In these experiments we were hand-tuning the conductances of excitatory synapses between DB motor neurons and muscle cells (only dorsal left). Because the results from the optimizer was not satisfactory, doing some hand-tuning allowed for the development of intuitions of the effects of changing certain parameters. In this case we are starting from only the parameters that provide a delayed spike time for the DB neurons, but not using the results of optimizing the spike frequencies from the 7-25 experiment. The target for the hand tuning is an improved version of having multiple spikes staggered without the jitter that was found in the 7-25 experiment. The strategy to begin with was to begin picking connections between motor neurons and muscles and changing conductances in a trial and error approach.

In case01-case05 we made some minor changes of the delayed GJs AVB-DB with the resulting parameters of a previous optimization, we also changed the conductance to muscles.
Although we could observe a wave of action potentials propagating from neck to tail (case05), there was an issue with these parameters.

We were optimizing the wave of the first spikes and after we increased the simulation duration we noticed that there was only one complete wave (case06) because motor neuron activity still changed after the first spike had occured. That meant there was jitter. Looking at the problem, the motor neurons don't have steady state behavior until later in the simulation.

In case07-case16 we were playing with an increased conductance of muscle-muscle GJs to try to improve the alignment of spikes

From case 16 on we were trying to create a wave propagating from neck to tail but this time we only had a look at the muscle output after motor neurons had reached steady state. This is done by changing the conductance between motor neuron and muscle cell. If there is a muscle spiking too early or too late for the sequence, then we are shifting it by changing the weight of the synaptic conductance from DB to muscle. In order to take into account the jitter in later spikes, instead of targeting the first spike time, we are looking at the timing of the second, third or fourth spike and adjusting the conductance to get it to be at the right time in seqence.

In case 30 -- they are well aligned in the DL muscles. The DRs are not optimized so they act as a control by using the default values. Increasing perfection was happening after that. By case 38 they are very well aligned, but the timings are not as linear as desired.

Network:

Neural Circuit:

• Command neurons: AVBL, AVBR
• Motor neurons: DB1, DB2, DB3, DB4, DB5, DB6, DB7
• Muscles: All

Connections:

• Gap-junctions:
  • AVBL - AVBR
  • DB1 - DB2, DB2 - DB3, ..., DB6 - DB7
  • muscle - muscle
• Delayed gap-junctions:
  • AVB(L/R) - DB(1-7) (no connection between AVBL - DB1)
• Chemical synapses:
  • AVBL - AVBR, AVBR - AVBL
  • AVBR - DB4
  • AVBR - MVL16
  • DB - muscle

Simulation Setup:

• duration: see experiment
• dt: 0.05
• Injected current:
  • Target: AVB(L/R)
  • Current: 15 pA
  • Time: 50 ms <= t < (duration - 100) ms

Result of experiments

Id	Neurons (membrane)	Neurons (membrane)	Muscles (membrane)	Muscles (membrane)	Neurons (activity)	Neurons (activity)	Muscles (activity)	Muscles (activity)
Case38
Case37
Case36
Case35
Case34
Case33
Case32
Case31
Case30
Case29
Case28
Case27
Case26
Case25
Case24
Case23
Case22
Case21
Case20
Case19
Case18
Case17
Case16
Case15
Case14
Case13
Case12
Case11
Case10
Case09
Case08
Case07
Case06
Case05
Case04_2
Case04
Case03
Case02
Case01