This page is a list of online tools to search for Gen1, VT, and Split-GAL4 driver lines that label a cell of interest. Usually, the neuron(s) are found in one of the connectome datasets, and their EM skeleton can be compared to existing genetic lines. This can be done through two main web pages, NeuronBridge and PatchPerPix. These two sites will output lines that MAY contain your neuron of interest. From there, these lines should be validated using FlyLight.

NeuronBridge#

NeuronBridge (Clements 2020) is a useful online search tool for identifying Gen1, VT, and Split-GAL4 driver lines for targeting a given neuron of interest. In general, one should use NeuronBridge to determine “are there driver lines that contain my neuron of interest”? For the question of “is my neuron of interest expressed in said driver line”, one should then turn to FlyLight (see last section). This is because the relative brightness of neurons labelled via MCFO/cMIP does not correlate with the relative brightness of the same neuron labelled with GAL4/UAS.

There are two ways to search NeuronBridge. One is by using the hemibrain ID of your neuron of interest. For this type of search, NeuronBridge now has two modes available. The first is “ColorDepth”, this method is better at digging through sparser MCFO images but may miss your neuron in denser ones. The second is “PatchPerPixel”, this method is better at digging through denser MCFO images, but can sometimes be harder to interpret by eye. When searching for lines, you should consider using both in the above order.

However, this method does not work as well for neurons that are not well represented in the hemibrain dataset. If this is the case, you can upload a ColorMIP file of the entire neuron to match with light-level data. A ColorMIP file is a 2 dimensional projection of your neuron, where the color corresponds to the depth in the brain. The first thing you will need is a nrrd file of your EM skeleton. This can be done in R using the flywire_to_nrrd() function developed by Alex and Emily (see here). Then, these skeletons can be transformed into a ColorMIP image using the ColorMIP_Mask_Generator (Otsuna & Kawase, 2018). Follow their READ_ME file for taking the nrrd files and transforming them to colorMIP images through a Fiji macro. Then, you can upload these to NeuronBridge and run only the “ColorDepth” method of searching.

Using NeuronBridge#

To start searching for LM matches for a given hemibrain neuron of interest, either click the NB icon directly in NeuPrint or enter the neuron ID on NeuronBridge:

You will then select your search algorithm as “Color Depth” or “PatchPerPix”

Or, click upload and drag your ColorMIP image.

Then, set your settings to search for light microscopy libraries, and I have found that the standard settings work just fine.

Through either method of searching, NeuronBridge will then give you a (likely very long) list of possible LM matches, ranked according to their pixel-by-pixel comparison. For most neurons, the match scores drops after the first ~10-20 matches. That said, the match scores are not perfect, so it is still worth your time to click through the entire list to ensure a good hit isn’t lurking somewhere down the line.

This can be done quite quickly by clicking on a hit, using the mirrored cursor to compare morphology between the neuron mask and search result, and using the left-right arrow keys to navigate down the list of hits.

PatchPerPix#

For a different interface, you can also search through PatchPerPix directly (Mais et. al. 2021), which is a better algorithm for especially large neurons or especially dense MCFO images.

Using PatchPerPix#

To search, paste in the hemibrain IDs of your neuron(s) of interest. Searching will give you a spreadsheet with the top 150 line names, ranks, and slide codes.

There will also be a PDF showing the line hits, MCFO, and overlayed skeleton and light level data. See the instruction page for more detail.

Evaluating a Possible Hit#

When you find a promising search hit, there are a few helpful tools and tricks for evaluating whether it is actually a good hit. Generally speaking, matches for your neuron on NeuronBridge/Flylight should be quite good. While the precise location of the cell body and primary neurite may be subject to displacement due to how the brain is mounted, the dendritic and axonal arbor morphology should be well preserved. In other words, if you aren’t sure whether or not the morphology looks like your neuron of interest, it is probably not your neuron of interest.

IMPORTANT: DO NOT JUDGE THE LINE BASED ON THE BRIGHTNESS OF THE MCFO CMIP IMAGE!!! While your cell of interest may look bright in MCFO, that does not mean that it will look bright when labelled under GAL4/UAS-GFP or UAS-jGCaMP! Finding matches on NeuronBridge is the first step, but you should not consider the hit a “useable candidate” until you have validated the expression. (However, in some cases the neuron may be faint within FlyLight images and bright in our own hands, but this seems to be more rare).

  • Check all MCFO MIP images - click on the “Line Name” in your search result window (blue arrow) or enter the line name in the main search bar to pull up all LM images available for that driver.
    • This can be really helpful, as different MCFO slide might give you better/brighter/less obstructed views of your neuron of interest.
    • This can also give you a sense of what other neurons are labeled by this line (messy or specific?).
    • HIGHLY RECOMMEND: for an MCFO slide that you think contains your neuron of interest, click “View EM Matches” on the righthand side. This will pull up all of the hemibrain neuron matches for that specific image.

  • Check FlyLight MCFO stacks - under External Links, click on “FlyLight Gen1 MCFO” to look through the non-compressed MCFO stacks for both the brain and VNC, or you can click on Virtual Fly Brain to look at the full expression pattern.
    • Note that the slide codes in NeuronBridge match those in FlyLight, so you can compare the MCFO stack with the color-MIP projection.
    • You can also download the original LSM files for each stack on this page.
    • HIGHLY RECOMMEND: It can sometimes be much easier to discern the morphology of individual colorized neurons in the MCFO stack rather than using the compressed MIP.
    • You should not rely on a particular line as being a good “hit” unless you’ve checked the FlyLight MCFO stacks.

Validate with FlyLight#

This is a really IMPORTANT final step! While your cell of interest may look robustly labeled using MCFO, that does not mean it will be brightly labelled when using a UAS or LEXAOP reporter gene. That is to say that many of your hits on NeuronBridge are, practically speaking, “false positives”, where the cell of interest is truly included in a given driver hit, but that driver does not express sufficient GAL4 in the cell of interest to use typical ephys/imaging reporters. In my experience, the MCFO images can actually be incredibly misleading when it comes to trying to judge the relative brightness of one cell or another. So if you want to know whether your line is ACTUALLY good, you need to look at the FlyLight full stack and download the LSM files (use 7-zip to open BZ2s).

  • To look at the reporter expression of your cell relative to every other cell contained in the driver, you should use the FlyLight GAL4/LEXA collection to lookup EVERY driver hit that you get from Neuronbridge.
  • When evaluating the expression pattern, consider:
    • Can you identify your cell?
    • Is your cell bright or dim relative to other cells that are labelled?
      • If your cell is hard to find or dim, it will likely be dim for imaging/patching
      • Score each hit based on how brightly your cell is labelled, ideally you should only be using the brightest lines possible
    • Are there any “distractor” cells near yours that would make it hard to identify for imaging/ephys?

Finding LexA lines#

Unfortunately, the NeuronBridge library only contains the expression patterns for GAL4 lines and not LEXA lines. For searching for LEXA lines, you should download the R/VT library aand use the manual cMIP search in FIJI.

Backward FlyWire Tracing from confocal images Forward