Lists of Neurons#

This tutorial will show you how to use NeuronLists to efficiently work with many neurons at a time.

NAVis will typically collect multiple neurons into a navis.NeuronList as container. This container behaves like a mix of lists, numpy arrays and pandas dataframes, and allows you to quickly sort, filter and manipulate neurons.

Overview#

import navis

# Grab three example skeletons (TreeNeurons) as a NeuronList
nl = navis.example_neurons(n=3)
nl

df = nl.summary()

import matplotlib.pyplot as plt

navis.plot2d(nl, view=('x', '-z'), method='2d')
plt.tight_layout()

Creating NeuronLists#

To create a NeuronList from scratch simply pass a list of neurons to the constructor:

n = navis.example_neurons(n=1)
nl = navis.NeuronList([n, n, n])  # a list with 3x the same neuron
nl

Accessing Neuron Attributes#

NeuronLists give you quick and easy access to data and across all neurons:

# Get the number of nodes in the first skeleton
nl = navis.example_neurons(n=3)
nl[0].n_nodes

Out:

4465

Use the NeuronList to collect number of nodes across all neurons:

nl.n_nodes

Out:

array([4465, 4847, 4332])

This works on any neuron attribute:

nl.cable_length

Out:

array([266476.88, 304332.66, 274703.38], dtype=float32)

If the neuron attribute is a dataframe, the NeuronList will concatenate them and add a new column with the neuron ID:

nl.nodes  # note the `neuron` column

NeuronLists can also contain a mix of different neuron types:

nl_mix = navis.example_neurons(n=2, kind='mix')
nl_mix

Note how nl_mix contains a TreeNeuron and a MeshNeuron?

In such cases you have to be a bit more careful about asking for attributes that are not shared across all neurons:

# This will throw an error because MeshNeurons
# don't have a `cable_length` attribute
nl_mix.cable_length

# Instead use the `get_neuron_attributes()` method with a default value:
nl_mix.get_neuron_attributes('cable_length', None)

Out:

array([266476.88, None], dtype=object)

Indexing NeuronLists#

A NeuronList works similar to normal lists with a bunch of additional perks:

# Get the first neuron in the list
nl = navis.example_neurons(n=3)
nl[0]

Index by position#

NeuronLists are designed to behave similar to numpy arrays in that they allow some fancing indexing.

You've already seen how to extract a single neuron from a NeuronList using a single integer index. Like with numpy arrays, this also works for lists of indices...

nl = navis.example_neurons(n=3)
nl[[0, 2]]

... or slices

nl[:2]

Index by attributes#

You can index NeuronLists by boolean numpy.arrays - that includes neuron attributes, e.g. n_nodes, cable_length, soma, etc.

Index using node count:

subset = nl[nl.n_branches > 700]
subset

Here is an example where we subset to neurons that have a soma:

subset = nl[nl.soma != None]  # Index by boolean array
subset

Index by name#

navis.TreeNeuron can (but don't have to) have names (.name). If you, for example, import neurons from SWC files they will inherit their name from the file by default.

Our example neurons all have the same name, so to demo this feature we will need to make those names unique:

for i, n in enumerate(nl):
    n.name = n.name + str(i + 1)
nl

You can index by single...

nl["DA1_lPN_R1"]

... or multiple names:

nl[["DA1_lPN_R1", "DA1_lPN_R2"]]

Using regex#

Under the hood NAVis uses re.fullmatch to match neuron names - so you can use regex!

nl[".*DA1.*"]

Index by ID#

All neurons have an ID - even if you don't explicitly assign one, a UUID will assigned under the hood.

nl[0].id

Out:

1734350788

Neuron lists can be indexed by their ID (similar to .loc[] in pandas DataFrames) by using the .idx indexer:

nl.idx[1734350908]

Neuron Math#

NAVis implements a very simple and intuitive syntax to add and remove items from a navis.NeuronList:

Addition#

To merge two lists in Python, you can simply add them:

[1] + [3]

Out:

[1, 3]

navis.NeuronList works exactly the same:

nl[:2] + nl[2:]

This also works on with two single navis.TreeNeurons! You can use that to combine them into a list:

nl[0] + nl[1]

Substraction#

To remove an item from a Python list, you would call the .pop() method:

l = [1, 2, 3]
l.pop(2)
l

Out:

[1, 2]

For navis.NeuronList you can use substraction:

nl - nl[2]

Bitwise AND#

To find the intersection between two lists, you would use sets and the & operator:

set([0, 1, 2]) & set([2, 3, 4])

Out:

{2}

navis.NeuronList work similarly:

nl[[0, 1]] & nl[[1, 2]]

Bitwise OR#

To generate the union between two lists, you would use sets and the | operator:

set([0, 1, 2]) | set([2, 3, 4])

Out:

{0, 1, 2, 3, 4}

navis.NeuronLists work similarly:

nl[[0, 1]] | nl[[1, 2]]

Multiplication and Division#

So far, all operations have led to changes in the structure of the navis.NeuronList. Multiplication and division are different! Just like multiplying/dividing individual neurons by a number, multiplying/dividing a navis.NeuronList will change the coordinates of nodes, vertices, etc. (including associated data such as radii or connector positions) of the neurons in the list:

nl.units  # our neurons are originally in 8x8x8 nm voxels

[8 8 8]

nl_um = nl * 8 / 1000  # convert neurons: voxels -> nm -> um
nl_um.units

[1.0 1.0 1.0]

The above will have changed the coordinates for all neurons in the list.

Comparing NeuronLists#

navis.NeuronList implements some of the basic arithmetic and comparison operators that you might know from standard lists or numpy.arrays. Most this should be fairly intuitive (I hope) but there are a few things you should be aware of. The following examples will illustrate that.

In Python the == operator compares two elements:

1 == 1

Out:

True

2 == 1

Out:

False

For navis.TreeNeuron this is comparison done by looking at the neurons' attribues: morphologies (soma & root nodes, cable length, etc) and meta data (name).

nl[0] == nl[0]

Out:

True

nl[0] == nl[1]

Out:

False

To find out which attributes are compared, check out:

navis.TreeNeuron.EQ_ATTRIBUTES

Out:

['n_nodes', 'n_connectors', 'soma', 'root', 'n_branches', 'n_leafs', 'cable_length', 'name']

Edit this list to establish your own criteria for equality.

For NeuronList, we do the same comparison pairwise between the neurons in both lists:

nl == nl

Out:

True

nl == nl[:2]

Out:

False

Because the comparison is done pairwise and in order, shuffling a NeuronList will result in a failed comparison:

nl == nl[[2, 1, 0]]

Out:

False

Comparisons are safe against copying but making any changes to the neurons will cause inequality:

nl[0] == nl[0].copy()

Out:

True

nl[0] == nl[0].downsample(2, inplace=False)

Out:

False

You can also ask if a neuron is in a given NeuronList:

nl[0] in nl

Out:

True

nl[0] in nl[1:]

Out:

False

Operating on NeuronLists#

With very few exceptions, all NAVis functions that work on individual neurons also work on navis.NeuronList.

n1, n2 = navis.example_neurons(2)  # grab two individual neurons

# This will raise an error
navis.downsample_neuron([n1, n2], 2)

# This will work
navis.downsample_neuron(navis.NeuronList([n1, n2]), 2)

NeuronList methods#

Similar to individual neurons, navis.NeuronLists have a number of methods that allow you to manipulate the neurons in the list. In fact, (almost) all shorthand methods on individual neurons also work on neuron lists:

nl = navis.example_neurons(2)
for n in nl:
   n.reroot(n.soma, inplace=True)  # reroot the neuron to its soma

nl = navis.example_neurons(2)
nl.reroot(nl.soma, inplace=True)  # reroot the neuron to its soma

In addition navis.NeuronLists have a number of specialised methods:

nl = navis.example_neurons(3)  # load a neuron list
df = nl.summary()  # get a summary table with all neurons
df.head()

# Quickly map new attributes onto the neurons
nl.set_neuron_attributes(['Huey', 'Dewey', 'Louie'], name='name')
nl.set_neuron_attributes(['Nephew1', 'Nephew2', 'Nephew3'], name='id')
nl

# Sort the neurons by their name
nl.sort_values('name')  # this is always done inplace
nl

Of course there are also a number of NeuronList-specific properties:

• is_mixed: returns True if list contains more than one neuron type
• is_degenerated: returns True if list contains neurons with non-unique IDs
• types: tuple with all types of neurons in the list
• shape: size of neuronlist (N, )

All attributes and methods are accessible through auto-completion.

What next?#

Total running time of the script: ( 0 minutes 3.210 seconds)

Download Python source code: plot_02_neuronlists_intro.py

Download Jupyter notebook: plot_02_neuronlists_intro.ipynb

Gallery generated by mkdocs-gallery