NBLAST

NBLAST is described in an open access manuscript Neuron: Costa et al. NBLAST: Rapid, sensitive comparison of neuronal structure and construction of neuron family databases published in Neuron. There is a companion website at http://bit.ly/nblast. You can look at NBLAST results via the Virtual Fly Brain website and examine the source code using the links described below.

Numerous large image datasets of single neuron or sparse genetic labelling data for the fruit fly brain have become available in recent years. As the scale and quality of these data have increased, 3D querying of data has become an increasingly relevant and potentially valuable application. We have developed an approach based on a simplified representation of neuron structure (very similar to Masse et al 2012), a simple 3d alignment approach based on 3d registration and nearest neighbour point search and a scoring system directly inspired by BLAST.

Although the approach is generic, we have carried out most of our testing on raw data kindly provided by http://flycircuit.tw. This image data was registered using CMTK against a newly constructed template brain, FCWB.

For the paper we carried out a complete clustering of the 16129 neurons from the flycircuit.tw dataset. This can be viewed online here, including interactive 3D rendering of clusters powered by WebGL.

We also provide the ability to generate the figure panels used in the paper. Instructions can be found here and a link to a video demo is provided below.

A simple web-app for comparing neurons using our NBLAST algorithm is available at http://jefferislab.org/si/nblast/on-the-fly/, which provides the ability to compare FlyCircuit neurons against one another and against user-uploaded tracings, with interactive 3D display, from the comfort of your browser.

• Looking at 3D rendering of NBLAST clusters via searches on the Virtual Fly Brain website by neuropil they overlap
• Online pairwise NBLAST search for two flycircuit neurons
• Generating a figure report using knitr and RStudio.
• Finding the most similar FlyCircuit neurons to a GAL4 tracing using RStudio and the R packages detailed below.

• Core Registration and munger wrapper script from CMTK
• Fiji/ImageJ
• unu
• R
• Scripts for template brain construction
• Coordination of registration/skeletonisation by nat.as

• All analysis was carried out in R using packages released as the Neuroanatomy Toolbox
• Core functions including data I/O and 3d plots are provided by the package nat
• Neuron search is implemented in nat.nblast
• Metadata and cached searches for the flycircuit.tw dataset are implemented in a dedicated flycircuit package.
• the full NeuroAnatomy Toolbox can be installed within an R session with

  if(!require("devtools")) install.packages("devtools")
  devtools::source_gist("39a1182f726989db7e03")

  (You can see the source for this gist here: https://gist.github.com/jefferis/39a1182f726989db7e03).