Caenorhabditis elegans (C elegans) Hermaphrodite Wild Type has a fully mapped connectome of 302 Neurons. C elegans is the most studied animal on earth.
The connections and connection types (Chemical and Electrical), and the function of each neuron is well known (Sensory, Interneuron and Motor Neurons)
Nervous systems are messy, noisy systems, each neuron works independently of one another, and I wanted to recreate that system as close as possible.
We have this vision that neurons are nice, neat tubes but in reality, they are like a play-doh snake someone has smashed and squished.
I created a single program that can assume the identity of each of the 302 neurons that make up the C elegans connectome.
Each of the 302 individual programs message to the other programs via User Datagram Protocol (UDP). UDP is like TCP but doesn’t care about packet loss. UDP is how most video and audio is transmitted over the internet.
To simulate connections, I used the number of connections as a weighted value so that if neuron A has 3 synaptic connections to neuron B, I send to neuron B a UDP message that contains a value of 3 when the action potential of A is reached.
Each neuron was given the same threshold value and when the incoming weighted values were summed, if the accumulated values were greater than the threshold, the neuron would message or send weighted values the neurons it connects too.
If 200ms passed between incoming messages, the accumulator was set to zero and the weighted accumulation would start all over again (Temporal aspect).
Each time a Neuron “Fires”, the accumulator is set to Zero (0).
GABAergic connections are assumed to be Inhibitory and so I used negative weights for these connections.
A Framework was created that allowed the sensor input from the robot to be read and stimulate the associated sensory neurons.
Sonar stimulated Nose Touch sensory neurons
Sound to stimulate Food sensory neurons.
Touch sensors for Anterior and Posterior body sensory neurons.
Likewise, the Framework was extended to capture the motor neuron weights to the individual body muscles.
The Dorsal and Ventral, Left and Right side body muscles were accumulated and the totals were used to determine the speed to drive the wheels on the robot.
The entire system is mapped as:
What did we learn?
It's rare to be able to do a whole organism simulation of any type.
The Connectome alone created the behaviors observed in the robot and these behaviors were similar to the biological worm. This was not a program that told the robot to stop, backup and go a different direction, only how the artificial neurons are wired. Well documented ablation experiments also showed similar, observable behaviors.
The connectome is always on. I did a test and after initial stimulation, the connectome continued to run for 24 hours and if I hadn’t stopped it, probably would have gone on forever.
The connectome is highly recurrent and further analysis shows that the connectome is exponentially recurrent.
Each neuron connects to a set of other neurons and as they are stimulated, a subset will connect back to the originating neuron. This connection back to the originating neuron grows exponentially the deeper the stimulation.
Beyond the C elegans simulation, we find more about connectomes.
There are two (2) basic temporal functions in the connectome:
The depolarization timing where if a neuron, real or simulated, is left alone or not stimulated over a certain timespan, it depolarizes or goes to a negative (or zero) value = internal function of time; i.e. the all or none response.
The network topology of the connectome has is a function of time as well whereas if certain neurons fire before others, a set of behaviors will occur but if a set of those same neurons fire in a different sequence, we often see other behaviors = an external, network timing.
Small incremental changes at the neuronal level is insignificant at the muscle output level; i.e. weights can accumulate very quickly between neurons so in a quick instance, the values can be high and then low, then high again, etc but the nature of the motor/muscle output averages this out as a function of time.
There is a quadlateral component (in C elegans it is delineated by Left and Right, Dorsal and Ventral) to connectomes that is very important in balancing connections which isn’t completely clear at high levels but for C elegans, it is a key component for the sinusoidal motion.
In C elegans, we see on average that for any given neuron with a Left and Right pair (e.g. interneuron AVB = AVBL and AVBR), for the connections those neurons have to other neurons, about 1/3 cross connect to the same neurons. As an example if neuron AVBL connects to 20 other neurons and neuron AVBR connects to 21 other neurons, they both connect to approximately 7 of the same neurons.
From other connectome analysis like Drosophila, we know that the recurrency is very strong in Local Cortical Modules and still present but weaker in the connections between Local Cortical Modules.
(Ref: Three-Dimensional Reconstruction of Brain-wide Wiring Networks in Drosophila at Single-Cell Resolution, Ann-Shyn Chiang, Chih-Yung Lin, et al)
What this experiment IS NOT
I did not take into consideration the Systems Biology and intracellular activity. Neuropeptides, Innexins, Synapse and Gap Junction differences, various neurotransmitters (excitatory and inhibitory classes only), and the various Receptors have been discounted.
Likewise, I do not take into consideration the ion channel differences and their complexities (Na+, K fast, K slow, Ca+, Cl-, etc); i.e. both Neuron and Muscle.
All Polarization and Depolarization timing across the system is the same.
There is no spatial consideration whereas some neurons run the entire length of the worm’s body and I treat all neuronal interaction the same spatially.
This experiment is purely Connectomics!
Comments
Post a Comment