Thanks to those of you who joined the Town Hall! While the video is still processing, I wanted to share a few followup links and notes here:
Re: @AzureJay question about authorship, the full author list will be so long that we will create a FlyWire Consortium (similar to the first Eyewire paper where we listed The Eyewirers* and then listed usernames in supplementary information). We’re not sure what format the journal will require yet but likely you will get to choose between username or full name.
As we approach completion of the connectome, annotations are becoming increasingly valuable. If you recognize cells that have not been labeled, even if you know general type vs specific, it would be helpful if you add annotations.
If there is interest, we could create a spreadsheet of citizen science completion history so that you could easily find cells that may be easy to identify. We’re also soliciting requests from researchers for types of neurons that they need annotations for. There is a new tool, CODA (Connectome Data Annotator) in the works, not sure when that will come out, but hopefully will make things faster and easier to annotate.
There’s a typo in my name on slide 32, lol (it should be Krzysztof, instead of Krzystof. Btw, if anyone wonders how to pronounce it, it’s something like Kshyshtof).
The Codex looks amazing. I’m looking forward to play with it.
I probably should concentrate more on labeling cells. For now I was mostly only proofreading and completeing cells, thinking, that the annotating could be done, after we finish all the neurons. I didn’t thought, that things (researching) can be done also on partial datasets.
Oh my gosh I’m so sorry! Corrected and I’ll use that pronunciation moving forward
Yes the annotations will be super useful, immediately. Thank you. Would it be helpful if we made some spreadsheets with completed OL neurons that need annotations?
If there are any specific neurons, that need to be annotated ASAP, then a spreadsheet could be useful.
Otherwise, at least for me, the Cell completion stats will give me work for quite a long time (I wil be annotating all the cells completed by me using the links in that topic).
This was a fascinating town hall and while a bit of it did go over my head (re: synapses) I’m still thrilled to be a part of this.
I would definitely be in support of a spreadsheet or even just a pre-saved search link for cells needing annotations that we could work through.
Amy, you also mentioned collaborations during your presentation and I would like to say I’d be glad to collaborate with someone! I have done some investigation of Tm and TmY typing alongside @TR77 (they have done far more than I have), however I am not specifically inclined and I’m happy to jump on partnership on any specific subset of data that might be needed. Otherwise, I might just pluck up a subset of cell types to start working on in general as a farm for proofreading and annotation.
is there any way to get multiple cell up at the same time from the list of completed cells,
when tracing it is often easy too see that it is probably a TM cell but much more difficult to see that it is a TM7 for instance. if we could get a batch of maybe 100 cells already marked as TM we could then pick out all identical cells and open them in a few new tabs and save links from each tab even if not knowing what specific type it is, it would be easier to use time to classify 10- 20 identical cells at once instead of doing it one at a time.
a spreadsheet would probably be a easy way to set in link with groups as a start and mark that someone is working on splitting that group. When the most dominant types is classified you could then merge two or more big groups so the classified groups not get too small
the same metod could probably be used to pick out wrongly classified cells or to find primary type on cells not yet classified
Does the lightbulb change if a cell has been classified? I was going to go back through my history but it’d be easier if I could tell at a glance if a given one needed classification.
if you mean if it has been given a cell type, no it will remain green coloured for completed. To find out -as of when this reply is written- you need to go click the lightbulb and then click cell identification.
For what it’s worth, I’ve decided to focus my efforts in cells with their soma outside/at the far edges of the lobula plate - T4, T5, Tlp, and Y types primarily.
We just got off a call with a researcher in France who is interested in the T4 cells. There are ~3,200 per optic lobe! He mentioned something similar to @annkri 's post about how it is much easier to tell subtypes (T4a, b, c, d) when there are many examples, so perhaps a first pass of broader class would pave the way for more specific annotations.
The annotation system in FlyWire has a lot of room for improvement so unfortunately yeah spreadsheets may be ideal at this time. Codex is static and only updated periodically (currently ~1x/mo, but should be more frequent) so the only way to see when annotations are added in real time is the Cell Identification page via lightbulb.
I got a good start this weekend with the T4/T5 cells in the area I’m looking through so I will continue to make that my focus! I have been keeping an annotation tab in neuroglancer with each soma I’ve checked and my ID on that soma, so I can also easily whip up a shareable view with T4s I’ve ID’d at any point.
Wow that’d be awesome! If you have a link like that, would you mind if I linked you up with Felipe over email?
I touched base with Sven and if you guys wanted to coordinate a batch of annotations via spreadsheet, we could ingest them into FlyWire. That way you wouldn’t have to click lightbulb menu 1,000x or whatever. The sheet would need: coordinates, segID, label. But of course you’re welcome to use lightbulb
in a way there is such a tab in the gsheet Q/A, “cell identification Q/A” if im understanding what you guys are discussing/wanting. the “identified” could be ‘it’? When we identify the/a cell we resolve it there and then at some point you guys (lab/HQ) integrate that info in fw? Can edit the tab to incl. the info you want like coords segid etc
i suspect it would be easier to just use the lightbulb 1,000x instead of manually writing down a bunch of seg id and cordinates, unless it is possible to export this for all cells in a tab or all cells in a annotation layer. Cordinates could be marked by annotations and would be faster to add than waiting on the lightbulb if you have many cells. labels would just be copy/paste.
@Nseraf i think we would probably need a new sheet with one tab for each main types of farm + a tab with cordinates, seg id and label for finished cells
Think i have identified 3 or 4 different T2 subtypes is this something we should make note of or just classify it as T2 regardless of the diffrances? FlyWire
Main differance is the axon part.
The red subtype might be two subtypes (think the cells with a second branch in the dendrites is probably what is labeled T2a)
(it should be Krzysztof, instead of Krzystof. Btw, if anyone wonders how to pronounce it, it’s something like Kshyshtof).
