Tutorials

• 1: Application Programming Interfaces (APIs)
• 1.1: Guide to Working with Images from Virtual Fly Brain (VFB) Using the VFBConnect Library
• 1.2: Hacking the connectome
• 1.2.1: Tool landscape
• 1.2.2: Introduction to connectomic data and tools
• 1.2.3: Discovery
• 1.2.4: Mapping
• 1.2.5: Visualisation
• 1.2.6: Connectomics
• 1.2.7: NBLAST
• 1.3: Downloading Images from VFB Using VFBconnect
• 1.4: VFB connect API overview
• 1.5: SOLR search example
• 1.6: Data types
• 1.7: Plotting
• 1.8: pymaid
• 1.9: NBLAST
• 1.10: neuprint

1 - Application Programming Interfaces (APIs)

1.1 - Guide to Working with Images from Virtual Fly Brain (VFB) Using the VFBConnect Library

Prerequisites

Before starting, ensure you have the VFBConnect library installed. The recommended Python version is 3.10.14, as this version is tested against the library.

pip install vfb-connect

Importing the VFBConnect Library

Start by importing the VFBConnect library. This library provides a simple interface to interact with neuron data from the Virtual Fly Brain.

from vfb_connect import vfb

Retrieving Neuron Data

To work with specific neurons, you can use the vfb.term() function. This function takes a unique identifier (e.g., ID, label, synonym) for the neuron.

Example: Retrieving a Single Neuron

neuron = vfb.term('5th s-LNv (FlyEM-HB:511051477)')

The neuron variable now holds data about the neuron identified by the given term.

Working with Neuron Data

The retrieved neuron object can provide different representations of neuron data, such as its skeleton, mesh, and volume. These representations can be visualized using various plotting methods.

# Access the skeleton representation
neuron_skeleton = neuron.skeleton

# Check the type of the skeleton representation
print(type(neuron_skeleton))  # Output: <class 'navis.neuron.Neuron'>

# Plot the skeleton in 2D
neuron_skeleton.plot2d()

# Access the mesh representation
neuron_mesh = neuron.mesh

# Access the volume representation
neuron_volume = neuron.volume

Retrieving Multiple Neurons

You can retrieve multiple neurons using the vfb.terms() function, which accepts a list of neuron identifiers.

Example: Retrieving Multiple Neurons

neurons = vfb.terms(['5th s-LNv', 'fru-M-300008', 'catmaid_fafb:8876600'])

This command retrieves multiple neurons, which can then be visualized or manipulated collectively.

Flexible Matching Capabilities

One of the key features of the VFBConnect library is its flexible matching capability. The vfb.terms() function can accept a variety of identifiers, such as:

• IDs: Unique identifiers assigned to each neuron.
• Xref (Cross-references): External references that relate to other datasets.
• Labels: Human-readable names for neurons.
• Symbols: Abbreviated names or symbols used to represent neurons.
• Synonyms: Alternative names by which a neuron might be known.
• Partial Matching: You can provide a partial name, and VFBConnect will attempt to find the best match.
• Case Insensitive Matching: Matching is case insensitive, so if an exact match isn’t found, ‘5th s-LNv’ and ‘5TH S-LNV’ are treated the same. This allows for more flexible querying without worrying about exact case matching.

Example: Using Flexible Matching

neurons = vfb.terms('5th s-LN')

If an exact match isn’t found, VFBConnect will provide potential matches. This feature ensures that even with partial or approximate information, you can still retrieve the relevant neuron data.

Output Example:

Notice: No exact match found, but potential matches starting with '5th s-LN': 
'5th s-LNv (FlyEM-HB:511051477)': 'VFB_jrchk8e0', 
'5th s-LNv': 'VFB_jrchk8e0'

This notice will help you identify the correct neuron based on the closest matches.

Visualizing Neurons

VFBConnect provides various methods to visualize neuron data, both individually and collectively.

3D Visualization

To plot neurons in 3D, use the plot3d() method. This is useful for visualizing the spatial structure of neurons.

neurons.plot3d()

2D Visualization

For 2D visualization, use the plot2d() method.

neurons.plot2d()

Viewing Merged Templates

VFBConnect also allows viewing merged templates of neurons, combining multiple neuron structures into a single view.

neurons.show()

Opening Neurons in VFB

To open the neurons directly in Virtual Fly Brain, use the open() method. This will launch a browser window displaying the neurons in the VFB interface.

neurons.open()

Summary

• Use vfb.term() to retrieve single neuron data.
• Use vfb.terms() to retrieve multiple neurons with support for partial, case-insensitive, and flexible matching (IDs, labels, symbols, synonyms, etc.).
• Access different data representations (skeleton, mesh, volume) via neuron objects.
• Visualize neuron data in 2D and 3D.
• Use the show() method to view merged neuron templates.
• Open neuron data directly in Virtual Fly Brain with the open() method.

These examples provide a foundation for working with neuron data from Virtual Fly Brain using the VFBConnect library. By exploring different neuron representations and visualization methods, you can analyze and understand neuron structures more effectively.

1.2 - Hacking the connectome

Below you can watch the recorded introduction session of our workshop and follow along with the workshop notebooks we work through to show examples of how you can use the available {{ ref “tools.md” tools }} to explore and analyse the published connectomes available on VFB.

1.2.1 - Tool landscape

Below are brief descriptions of the libraries/packages. For details, I defer to their respective (excellent) documentations.

Querying VFB

Queries against VFB’s REST API are easiest with vfb_connect for Python. For R there is a vfb_connect wrapper, vfbconnectr. See also David’s presentation for details.

R

In R, the natverse is your one-stop-shop for all things neuron: it’s a collection of various R packages that are built on top of the neuroanatomy toolbox, nat. Of particular relevance for this workshop:

1. nat is a general purpose library for working with morphological neuron data. In this workshop, we make heavy use of nat’s plotting capabilities but its capabilities extend far beyond that. If you want to run any morphological analysis, I highly recommend you have a look at the “Articles” in nat’s doc.
2. neuprintr and hemibrainr provide an interface with neuprint and the Janelia hemibrain dataset (link). The former lets you run queries against neuprint’s neo4j database while the latter contains meta data and various convenience functions to work with the hemibrain dataset.
3. rcatmaid provides an interface with CATMAID servers such as those the VFB uses to host published from the FAFB or larval fruit fly dataset. rcatmaid is built on top of nat and you can use nat functions with neurons pulled via rcatmaid.

Python

In Python, we find packages analogous to those in R:

1. navis is nat’s serpentine sibling: a general purpose neuron library for visualization and analysis of neuronal morphologies. It also features interfaces e.g. with Blender 3D and the natverse via rpy2.
2. python-neuprint is a Python library to interface with neuprint maintained by Janelia. Note that navis wraps this library and adds some convenience functions. See this tutorial.
3. pymaid lets you interface with CATMAID servers. Critically, it’s built on top of navis and you can natively use navis functions with pymaid neurons. Note that due to a name clash the library is called python-catmaid on PyPI.

Noteworthy mentions

There are a few more packages/functions that you might hear of over the course of the workshop.

NBLAST

NBLAST is an algorithm that computes morphological similarity between neurons (Costa et al., 2016). This has proven incredibly useful to find similar neurons across datasets but also to cluster neurons into cell types.

On the R side the algorithm is implemented in nat.nblast and in Python it is part of navis (see this tutorial).

Transforms

You will note that neurons pulled from VFB are typically in the same template space which makes co-visualization of neurons from different datasets a breeze. If you want to transform spatial data between template brains, e.g. from FAFB (“FAFB14”) to hemibrain (“JRCFIB2018F”), you should look for nat.flybrains & nat.jrcbrains in R and navis-flybrains in Python.

1.2.2 - Introduction to connectomic data and tools

1.2.3 - Discovery

Required packages: vfb-connect and python-catmaid (pymaid & navis)

!pip install vfb-connect --upgrade
!pip install python-catmaid --upgrade

A note on using these notebooks

This is designed as an interactive tutorial. Feel free to add code cells below each example to try out variations of your own.

How to find neurons across datasets

VirtualFlyBrain integrates images and connectomics profiles of neurons from many sources. It classifies and records their properties using a standard, queryable classification (The Drosophila Anatomy Ontology). This standardises the names of neuron types across sources, so you don’t need to worry about differences in nomenclature uses and supports queries for neurons by their classification.

# Import libs and initialise API objects
from vfb_connect.cross_server_tools import VfbConnect
import pandas as pd
vc = VfbConnect()

import pymaid
import navis

navis.set_pbars(jupyter=False)
pymaid.set_pbars(jupyter=False)

# Connect to the VFB CATMAID server hosting the FAFB data
rm = pymaid.connect_catmaid(server="https://fafb.catmaid.virtualflybrain.org/", api_token=None, max_threads=10)

# Test call to see if connection works 
print(f'Server is running CATMAID version {rm.catmaid_version}')

WARNING: Could not load OpenGL library.
INFO  : Global CATMAID instance set. Caching is ON. (pymaid)
Server is running CATMAID version 2020.02.15-905-g93a969b37

Finds neurons by type (classification) across datasets

We can use the vc.get_instances method in combination with the name of a neuron type on VFB to find individual neurons from multiple sources.

Use the search tool on VFB to find neuron types by name or synonym:

Use either the full name or the Symbol to query for neurons:

DA3adPN = vc.get_instances("adult Drosulfakinin neuron", summary=True)
pd.DataFrame.from_records(DA3adPN)

Find neurons by location

We can use the same method to search for neurons by location, using simple queries.

# Find neurons by location. The following query works across multiple data sources and both sides of the brain.  
# Results may be incomplete & may include minor overlap inferred from low synapse counts

neurons_in_DA3 = vc.get_instances("'neuron' that 'overlaps' some 'antennal lobe glomerulus DA3'", summary=True)
neurons_in_DA3_tab = pd.DataFrame.from_records(neurons_in_DA3)
neurons_in_DA3_tab[0:5]

Running query: FBbt:00005106 that RO:0002131 some FBbt:00003934
Query URL: http://owl.virtualflybrain.org/kbs/vfb/instances?object=FBbt%3A00005106+that+RO%3A0002131+some+FBbt%3A00003934&prefixes=%7B%22FBbt%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FFBbt_%22%2C+%22RO%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FRO_%22%2C+%22BFO%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FBFO_%22%7D&direct=False
Query results: 158

# Find local interneurons (intrinsic neurons) of the AL, overlapping DA3:

local_in_DA3 = vc.get_instances("'local interneuron of adult antennal lobe' that 'overlaps' some 'antennal lobe glomerulus DA3'",
                                summary=True)
local_in_DA3_tab = pd.DataFrame.from_records(local_in_DA3)
local_in_DA3_tab

Running query: FBbt:00007390 that RO:0002131 some FBbt:00003934
Query URL: http://owl.virtualflybrain.org/kbs/vfb/instances?object=FBbt%3A00007390+that+RO%3A0002131+some+FBbt%3A00003934&prefixes=%7B%22FBbt%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FFBbt_%22%2C+%22RO%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FRO_%22%2C+%22BFO%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FBFO_%22%7D&direct=False
Query results: 53

# Find neurons by dataset/paper - on CATMAID

bates = pymaid.find_neurons(annotations='Paper: Bates and Schlegel et al 2020')
bates

INFO  : Found 583 neurons matching the search parameters (pymaid)

# Inspect what datasets are available on VFB

ds = vc.neo_query_wrapper.get_datasets(summary=True)
ds_tab = pd.DataFrame.from_records(ds)
ds_tab.sort_values(by=['id'])

sayin_tab = pd.DataFrame.from_records(vc.get_instances_by_dataset('Sayin2019', summary=True))
sayin_tab

vc.get_connected_neurons_by_type(upstream_type='LNd', downstream_type='adult descending neuron', weight=20)

# Intra pacemaker neuron neuron connections

vc.get_connected_neurons_by_type(upstream_type='pacemaker neuron', downstream_type='pacemaker neuron', weight=20)

vc.get_connected_neurons_by_type(upstream_type='adult neuron', downstream_type='adult Drosulfakinin neuron', weight=20)

1.2.4 - Mapping

!pip install vfb-connect --upgrade

# Import libs and initialise API objects
from vfb_connect.cross_server_tools import VfbConnect
import pandas as pd

vc = VfbConnect()

import pymaid
import navis

navis.set_pbars(jupyter=False)
pymaid.set_pbars(jupyter=False)

# Connect to the VFB CATMAID server hosting the FAFB data
rm = pymaid.connect_catmaid(server="https://fafb.catmaid.virtualflybrain.org/", api_token=None, max_threads=10)

# Test call to see if connection works 
print(f'Server is running CATMAID version {rm.catmaid_version}')

# Many functions return JSON-compatible nested data structures. This function coverts them to DataFrame.
def summary_2_df(summary, sort=None):
    """Convert summary to DataFrame.  Optionally specify a set of columns to sort as a list of strings"""
    if sort:
        return pd.DataFrame.from_records(summary).sort_values(sort)
    else:
        return pd.DataFrame.from_records(summary)

WARNING: Could not load OpenGL library.
INFO  : Global CATMAID instance set. Caching is ON. (pymaid)
Server is running CATMAID version 2020.02.15-925-gf56795c9c

Mapping

Use case: If I have a single neuron, how can I find other neurons of the same or similar type within or between data sources?

Mapping via common parent type

# lets take some examples from a discovery query on the previous spreadhseet
sayin_tab = pd.DataFrame.from_records(vc.get_instances_by_dataset('Sayin2019', summary=True))
sayin_tab

oct_VPM3 = summary_2_df(vc.get_instances('octopaminergic VPM3 neuron', summary=True))
oct_VPM3

oct_VPM3_images = vc.neo_query_wrapper.get_images(oct_VPM3['id'], stomp=True, template='JRC2018Unisex', image_folder = 'oct_VPM3b')
oct_VPM3_images

oct_VPM3_images = vc.get_images_by_type('octopaminergic VPM3 neuron', stomp=True, template='JRC2018Unisex', image_folder = 'oct_VPM3')
nl = navis.read_swc('./oct_VPM3')
navis.plot3d(nl)

Running query: FBbt:00110151
Query URL: http://owl.virtualflybrain.org/kbs/vfb/instances?object=FBbt%3A00110151&prefixes=%7B%22FBbt%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FFBbt_%22%2C+%22RO%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FRO_%22%2C+%22BFO%22%3A+%22http%3A%2F%2Fpurl.obolibrary.org%2Fobo%2FBFO_%22%7D&direct=False
Query results: 3