Overview#

FlyWire is a platform for interacting with the FAFB dataset. There are some key differences relative to v14/CATMAID/walled garden:

  • The FlyWire team realigned the raw EM data, and this alignment is better than the v14 alignment.
  • As a result, the autosegmentation is better, with many fewer small fragments.
  • FlyWire uses the Neuroglancer interface rather than CATMAID.
  • FlyWire is open to everybody.

Dorkenwald et al. 2022 provides an overview of FlyWire.

The Flywire Quick Start Guide and the FAQ page provides basic instructions.

Check the social contract before you start using FlyWire. In particular, before you use any unpublished FlyWire data in a manuscript, talk, or poster, Rachel must get the approval of every PI whose lab identified those neurons or contributed more than 10% of the edits on those reconstructions. When we publish results from FlyWire, we should cite [Dorkenwald et al., 2020](doi: https://doi.org/10.1101/2020.08.30.274225).

Laia Serratosa Capdevila has written a quick and pragmatic introduction to using FlyWire:

FlyWire tracing — FlyWire tracing z=5284 (reload neurons on links for example solutions)

Brain regions#

You can view brain regions in FlyWire by navigating to this scene. You can then drag-and drop the green control-layer-rectangle into the FlyWire environment you are working in.

Locating neurons of interest#

There are several alternative approaches:

  • We can identify likely match to a cell type of interest by visually screening FlyWire neurons that occupy a subvolume-of-interest (often a primary neurite tract or axon tract) that must be occupied by the neuron of interest.
  • We can use NeuronBridge to identify a hemibrain neuron exemplar of our neuron-of-interest, starting with a manual tracing or a MCFO clone, etc. Then we can use NBLAST to find FlyWire candidates that seem to match this hemibrain neuron.
  • We can start with a hemibrain neuron and search potential FlyWire neurons via NBLAST. Ask Alex for help if you want to try this.
  • Or, with a light-level image stack, we can manually trace and register the brain to the FAFB template. From this, we can extract approximate FAFB coordinates of the neuron. We can then go into FlyWire and visually screen neurons around target coordinates to find our cell of interest. This works best when starting with coordinates of large-diameter neurites. 

Wilson lab catalog#

Once you have identified a neuron of interest, you should log it in the Wilson lab FlyWire Central Google Sheet. This allows other lab members to track your ongoing edits and annotations.

The following fields are required:

  • The user (initials). Rows will be automatically sorted overnight by user initials.
  • The flywire xyz coordinate of this neuron (copy in raw flywire voxel space).
  • The status field is important. You can put a single letter, and it will be updated to a full word. Please use only the following terms:
    • u = unassessed [not yet examined by a trusted human]
    • i = incomplete [small fragment]
    • c = complete [well fleshed out neuron, may even have most medium/small branches]
    • a = adequate [there is a cell body fibre, axon and dendrite]
    • m = merge error, noticeable merge error [this neuron is merged to another]
    • e = needs extending [not quite adequate, but more than a tiny fragment]
    • n = not a neuron [this segmentation is not a neuron, i.e. glia, erroneous] 

The initial status should be ‘unassessed’. If you think it is worth completing, then enter ’trace’ into the workflow to indicate that you plan to do some work on it.

Input/output searches#

Tools in Google sheets#

  • The workflow column in the Wilson lab FlyWire Central Google Sheet allows you to call several automated analyses. Note that these analyses should not be called until the neuron is deemed ‘complete’, i.e. its medium branches are approximately correct.
  • The automated analyses you can call are
    • i = inputs: this identifies input neurons
    • o = outputs: this identifies outputs to the neuron
  • To call multiple automated analyses, separate with a slash (e.g., i/o).
  • These commands will create new tabs. Each tab is labeled whimsy_inputs/outputs. The tracing tabs contain the ID for the neuron of interest (labelled as “main”) as well as a list of flywire IDs, their connection weights, their default status (unassessed). It does not contain flywire positions.
  • Each ID is updated nightly. You may find that some IDs are no longer valid. This happens when they have ‘changed ID’ due to a tracing update. Ignore these and move forward until the next day, when they will be updated/tossed automatically.
  • Each input/output is ranked according to its predicted synaptic weight. For standard input/output workflows, synaptic weights below a standard threshold have been cut off.
  • The status field is important. Please use the standardized options provided previously. The idea is for tracers to update statuses as they move down the list and reconstruct neurons to “adequate” or “complete” levels as they can.
  • Neurons identified through an input/output workflow are not added to the central spreadsheet. However, they can be added manually if you would like to assign them a workflow in subsequent steps.

Brain Circuits tool#

As an alternative approach, you can find candidate synaptic partners in FlyWire using Stephan Gerhard’s Brain Circuits tool. You can create an account at https://braincircuits.io/. (You might need to clear the cache in Google chrome by pressing Ctrl-Shift-Del, if you’re seeing a 404 error.) Note that neurotransmitter prediction in this site is for a given synaptic connection (and not averages across all the synapses for that segment ID). Note also that this site requires segment IDs from the “real” instance, not the “sandbox” instance.

Importing and viewing skeletons in FlyWire#

You can import skeletons and view them in FlyWire by navigating to https://flywiregateway.pniapps.org/. You can import any skeleton pre-defined in FlyCircuit and Hemibrain into FlyWire by just searching the IDs within this platform (skeletons can be mirrored to the other hemisphere within this tool). In addition, you can import a skeleton outside these databases as long as it is registered to any of the available reference atlases within this tool (e.g. FCWB, JRC2018U, etc).

Viewing hemibrain neurons#

You can view hemibrain neurons in FlyWire by navigating to this scene. You can then drag-and drop the green control-layer-rectangle (labeled hemibrain-meshes) into the FlyWire environment you are working in.

Backward FANC Searching for Genetic Lines Forward