Crate synapse_models

Source

Expand description

§Synapse Models Library

A comprehensive Rust library for modeling synaptic dynamics in computational neuroscience.

§Overview

This library provides detailed biophysical models of synaptic transmission including:

Neurotransmitter dynamics: Multiple neurotransmitter types (glutamate, GABA, dopamine, etc.) with realistic release and clearance kinetics
Receptor models: Detailed kinetic models for ionotropic (AMPA, NMDA, GABA-A) and metabotropic (GABA-B, mGluR) receptors
Vesicle pool dynamics: Tsodyks-Markram model for short-term depression and facilitation
Calcium dynamics: Pre- and postsynaptic calcium with buffering, stores, and CICR
Plasticity rules: STDP, BCM, Oja’s rule, Hebbian learning, homeostatic plasticity
Network models: Support for chemical synapses, gap junctions, ephaptic coupling, and neuromodulation

§Features

Biologically accurate parameters based on experimental data
Efficient numerical integration using exponential Euler methods
Thread-safe design for parallel simulations
Comprehensive test coverage
Extensive documentation with neuroscience background

§Example: Basic Synaptic Transmission

use synapse_models::synapse::Synapse;

// Create an excitatory synapse
let mut synapse = Synapse::excitatory(1.0, 1.0)?;

// Presynaptic spike at t=0
synapse.presynaptic_spike(0.0)?;

// Simulate for 10 ms
for t in 0..100 {
    let time = t as f64 * 0.1;
    synapse.update(time, -65.0, 0.1)?;

    // Get postsynaptic current
    let current = synapse.current(-65.0);
}

§Example: STDP Learning

use synapse_models::synapse::Synapse;

let mut synapse = Synapse::excitatory(0.5, 1.0)?;

let initial_weight = synapse.weight;

// Pre before post -> potentiation
synapse.presynaptic_spike(0.0)?;
synapse.postsynaptic_spike(10.0)?;

assert!(synapse.weight > initial_weight);

§Example: Network Simulation

use synapse_models::network::SynapticNetwork;
use synapse_models::synapse::Synapse;

// Create a network with 10 neurons
let mut network = SynapticNetwork::new(10);

// Add excitatory connections
for i in 0..9 {
    let syn = Synapse::excitatory(1.0, 1.0)?;
    network.add_connection(i, i + 1, syn)?;
}

// Spike from first neuron
network.spike(0)?;

// Update network
let voltages = vec![-65.0; 10];
network.update(&voltages, 0.1)?;

§Biophysical Background

§Synaptic Transmission

Synaptic transmission involves multiple steps:

Action potential arrival triggers voltage-gated calcium channels
Calcium influx causes vesicle fusion and neurotransmitter release
Neurotransmitter diffusion across the synaptic cleft (~20 nm)
Receptor binding opens ion channels or activates second messengers
Postsynaptic current flows, changing membrane potential
Neurotransmitter clearance by reuptake or degradation

§Short-Term Plasticity

Short-term plasticity operates on timescales of milliseconds to seconds:

Depression: Depletion of readily releasable vesicle pool
Facilitation: Residual calcium enhances release probability
Modeled by Tsodyks-Markram equations

§Long-Term Plasticity

Long-term plasticity underlies learning and memory:

STDP: Spike timing-dependent modification (±20-40 ms window)
LTP/LTD: Long-term potentiation/depression
Requires postsynaptic calcium elevation
CaMKII activation → LTP, calcineurin activation → LTD

§Mathematical Models

§Receptor Kinetics

First-order binding scheme:

dR/dt = α[NT](1-R) - βR

where R is open probability, [NT] is neurotransmitter concentration, α is binding rate, β is unbinding rate.

§NMDA Voltage Dependence

Mg²⁺ block (Jahr & Stevens, 1990):

B(V) = 1 / (1 + [Mg²⁺]/3.57 * exp(-0.062*V))

§Tsodyks-Markram Model

dx/dt = (1-x)/τ_rec - U*x*δ(t-t_spike)
du/dt = (U₀-u)/τ_facil + U₀(1-u)δ(t-t_spike)

§STDP Window

Δw = A₊ exp(-Δt/τ₊)     for Δt > 0 (potentiation)
Δw = -A₋ exp(Δt/τ₋)     for Δt < 0 (depression)

§Performance Considerations

Uses exponential Euler integration for numerical stability
Sparse network representation for efficiency
Minimal allocations in update loops
Thread-safe for parallel neuron updates

§References

Tsodyks & Markram (1997). The neural code between neocortical pyramidal neurons depends on neurotransmitter release probability.
Bi & Poo (1998). Synaptic modifications in cultured hippocampal neurons: dependence on spike timing, synaptic strength, and postsynaptic cell type.
Jahr & Stevens (1990). Voltage dependence of NMDA-activated macroscopic conductances predicted by single-channel kinetics.
Bienenstock et al. (1982). Theory for the development of neuron selectivity: orientation specificity and binocular interaction in visual cortex.

Re-exports§

pub use error::Result;
pub use error::SynapseError;
pub use synapse::Synapse;
pub use synapse::SynapseBuilder;
pub use synapse::SynapseType;
pub use network::SynapticNetwork;
pub use network::NetworkStats;

Modules§

calcium: Calcium dynamics in pre- and postsynaptic compartments.
error: Error types for synapse models library.
network: Network-level synaptic dynamics.
neurotransmitter: Neurotransmitter types and their biophysical properties.
plasticity: Synaptic plasticity rules for learning and memory.
receptor: Receptor dynamics and kinetic models.
synapse: Complete synapse model integrating all components.
vesicle: Vesicle pool dynamics and neurotransmitter release mechanisms.

Constants§

VERSION: Library version.