BigNeuron

Overview

Overview

Why do we need BigNeuron?

The three-dimensional shape of a neuron plays a major role in determining its connectivity, integration of synaptic inputs and cellular firing properties, and also changes dynamically with its activity and the state of the organism. Analyzing the three-dimensional shape of neurons in an unbiased way is critical to understanding how neurons function and developing applications to model neural circuitry.

The Problem

Advances in brain cellular imaging have now yielded thousands of detailed images of neurons from dozens of different organisms stored in personal collections across the globe, comprising many petabytes of data. Dozens of different imaging paradigms and algorithms have now been generated for visualizing the 3D structure of neurons from labs around the world. In order for large data sets to be cross-compared, however, the neuroscience field needs standards – for collecting the data, for determining the acceptable levels of resolution that are most suitable for analysis, and for deciding which analysis approach is the most effective for the most important questions driving the field.

The Solution

BigNeuron is a community effort to define and advance the state-of-the-art of single neuron reconstruction and analysis, and to create a common platform for analyzing 3D neuronal structures. The major goal of BigNeuron is to bench-test on a common open platform as many open-source, automated neuron reconstruction algorithms as possible with very large scale, publicly available single 3D neuron image data sets acquired by several light microscopy methods. The bench-test results will be produced using some of the fastest supercomputers in the world and then compared and validated using multiple criteria carefully defined by the computational neuroscience community.

The BigNeuron project has three phases.

  • In Phase I (2015) of the project, a series of hackathons will be held in Beijing (China), Cambridge (UK) and at Janelia (USA), cosponsored by a range of organizations. Neuron reconstruction algorithms contributed from participants will be ported onto a common software platform to analyze neuronal physical structure using the same core data set. Additional 3D images from multiple organisms will be solicited internationally during this time.
  • In Phase II (2015-2016), all ported algorithms will be bench-tested at DOE (USA) and Human Brain Project supercomputing centers, allowing us to standardize optimal protocols for labeling, visualizing and analyzing neuronal structure and key biological features. Computational annotation and analysis workshops will also be held at the Allen Institute (Seattle), Broad Institute (MIT) and in other ways, to help analyze the performance of neuron reconstruction.
  • In Phase III (2016) the combined results of Phases I and II will be used to develop a comprehensive annotated database of complex neuronal morphology, generate a searchable tool for discovering annotated and unique characteristics of neuronal morphology, and lay the groundwork for potentially integrating this with large-scale data sets linking form to neuronal function.

We will host events in different international locations to cross-compare and test the efficacy of analysis approaches on neurons of different known functions. We may also potentially crowd-source images. In so doing, we will develop the means to collectively develop a morphological pattern recognition database that will be an essential tool for understanding the individual components that sculpt brain circuitry.

A core group of collaborators have already performed significant preliminary work and provided critical resources to a pilot project that has been used to design this evolving effort. We plan to release a public call for algorithms and test data sets from worldwide researchers in March 2015, with the first data release from Phase I expected in early 2016.