This page was generated from docs/extension_multiflux.ipynb. Interactive online version: Colab badge Deepnote badge. Binder badge Download notebook.

NeuroML extension: Multiple flows per neuron

In the NeuroML formulation, neurons are stimulated by various synapses and input sources that inject non-specific current in them, influencing the membrane potential.

However, there may often be different flows of ions or other substances, that should be tracked separately of total electrical current. Or more generally, a neuron may be influenced by a sum of mechanisms that supply something other than current.

To cover this use case, EDEN allows mechanisms to inject multiple different things flowing into cells, like ion channels do already. This is done by an existing pure-LEMS concept, that is generalised in EDEN so that different mechanisms can provide different sets of flows.

Tip: This is similar to how pure LEMS works already, with the difference that not all mechanisms need to expose every sort of flow; each mechanism that doesn’t expose one of the additional flows is assumed to contribute 0 towards that flow.

Note: This feature is only supported for point neurons at the moment.

The <DerivedVariable> tag revisited

As we saw in the LEMS chapter, a <DerivedVariable> may take a value as a LEMS expression of other variables. Another option to resolve the value is by a select aggregate expression. EDEN supports the following tag attributes that describe the aggregate expression:

  • select as synapses[*]/, followed after the / by the name of an <Exposure> that some synapses or probes may offer.

  • reduce as add only, to add up the exposures of the same name.

The tag to gather all exposures of e.g. iB then gets to look like this:

<DerivedVariable name="ib" dimension="current" select="synapses[*]/iB" reduce="add" />

To prevent ambiguity, each flow must have the same <Exposure> name among all mechanisms that contribute to it. Hence different flows should be exposed by components as something other than i.

  • If components must expose the same flow under different names (or the modeller happens to treat them the same in one case), they can be select ed separately and an additional <DerivedVariable> may be defined for the total flow.

Note: Presently, the LEMS formulation of NeuroML concepts adds all “sort of external” mechanisms (known as Attachment>s) all into a singular synapses group, that includes both synapses instantiated by <connection>s and <input>s from <inputList>s!

An example of non-current input flow is seen in Tsodyks, Uziel, Markram (2000). More examples using different flows may be provided, as requested. See another usage example in EDEN’s program tests which shows how to separate the flow from different synapses and input probes.