NBLAST against FlyCircuit#

Match a query neuron against the entire FlyCircuit light-level dataset.

This example is not executed

In contrast to almost all other tutorials, this one is not executed when the documentation is built. Consequently, it also does not show any code output or figures. That's because this example requires downloading a large dataset (~850Mb) and running an NBLAST against it, it is simply not feasible to run this as part of the documentation build process.

If you work with Drosophila, chances are you have heard of FlyCircuit. It's a collection of ~16k single neuron clones published by Chiang et al. (2010).

For R, there is a package containing dotprops and meta data for these neurons: nat.flycircuit. This does not yet exist for Python. However, it's still pretty straightforward to run an NBLAST against the entire flycircuit dataset in Python!

First you need to download flycircuit dotprops from Zenodo (~850Mb). Next, unzip the archive containing the dotprops as CSV files. Now we need to load them into NAVis (adjust filepaths as required):

import navis
import pandas as pd

from pathlib import Path
from tqdm import tqdm

def load_dotprops_csv(fp):
    """Load dotprops from CSV files in filepath `fp`."""
    fp = Path(fp).expanduser()

    files = list(fp.glob('*.csv'))
    dotprops = []
    for f in tqdm(files, desc='Loading dotprops', leave=False):
        csv = pd.read_csv(f)

        pts = csv[['pt_x', 'pt_y', 'pt_z']].values
        vect = csv[['vec_x', 'vec_y', 'vec_z']].values

        dp = navis.Dotprops(points=pts, k=20, vect=vect, units='1 micron')
        dp.name = dp.id = f.name[:-4]

        dotprops.append(dp)

    return navis.NeuronList(dotprops)

For each CSV file (one neuron each), the loader does:

csv = pd.read_csv(f)                                                # (1)!

pts = csv[['pt_x', 'pt_y', 'pt_z']].values                         # (2)!
vect = csv[['vec_x', 'vec_y', 'vec_z']].values

dp = navis.Dotprops(points=pts, k=20, vect=vect, units='1 micron')  # (3)!
dp.name = dp.id = f.name[:-4]                                       # (4)!

Each CSV holds one neuron — read it into a DataFrame.
The pt_* columns are the point coordinates; the vec_* columns the associated tangent vectors.
Build a navis.Dotprops from the points and vectors (already in microns).
Use the filename (minus the .csv) as the neuron's name and id.

Load dotprops from the filepath

fc_dps = load_dotprops_csv('flycircuit_dotprops')
fc_dps

Generating these dotprops is slow, so pickle them once and reload from the pickle next time — much faster than parsing the CSVs again:

SaveLoad

import pickle

with open('flycircuit_dps.pkl', 'wb') as f:
    pickle.dump(fc_dps, f)

import pickle

with open('flycircuit_dps.pkl', 'rb') as f:
    fc_dps = pickle.load(f)

The names/ids correspond to the flycircuit gene + clone names:

fc_dps[0]

In case your query neurons are in a different brain space, you can use flybrains to convert them to flycircuit's FCWB space.

Requires flybrains

Converting between brain spaces needs the optional flybrains package and its bridging transforms (downloaded once via the flybrains.download_... functions).

For demonstration purposes we will use the example neurons - olfactory DA1 projection neurons from the hemibrain connectome - that ship with NAVis and try to find their closest match in the FlyCircuit dataset.

# Load some of the example neurons
n = navis.example_neurons(3)

# Convert from hemibrain (JRCFIB2018Fraw) to FCWB space
import flybrains

n_fcwb = navis.xform_brain(n, source='JRCFIB2018Fraw', target='FCWB')

A sanity check to make sure the transform worked

fig, ax = navis.plot2d([n_fcwb, flybrains.FCWB])

FlyCircuit neurons are all on the left side of the brain. We need mirror our neurons from the right to the left to match that.

n_fcwb_mirr = navis.mirror_brain(n_fcwb, template='FCWB')

Convert our neurons to dotprops

n_dps = navis.make_dotprops(n_fcwb_mirr, resample=1, k=None)

Run a "smart" NBLAST to get the top hits

scores = navis.nblast_smart(query=n_dps, target=fc_dps, scores='mean', progress=False)
scores.head()

Note

If you get a warning about data not being in microns: that's a rounding error from the transform and can be safely ignored.

Find the top hits for each of our query neurons

import numpy as np

for dp in n_dps:
    hit = scores.columns[np.argmax(scores.loc[dp.id].values)]
    sc = scores.loc[dp.id].values.max()
    print(f'Top hit for {dp.id}: {hit} ({sc:.3f})')

All of our query neurons should have the same top match (they are all from the same cell type after all): FruMARCM-F001496_seg001

Let's co-visualize: Queries in red, hit in black

fig, ax = navis.plot2d([n_fcwb_mirr, fc_dps.idx['FruMARCM-F001496_seg001']],
                       color=['r'] * len(n_fcwb_mirr) + ['k'])