(Almost) all cells, that have synapses in the left medulla

#1

For the last week I’ve been working on this little project, where I’ve found a group of wide-field neurons, that have synapses only or mostly in the medulla. My goal was to completely cover each of the medulla’s layers.
The cells, I’ve chosen, are here (thanks @st0ck53y for all the Dm4s):

https://ngl.flywire.ai/?json_url=https://globalv1.flywire-daf.com/nglstate/5357095591346176

The next step was to use the Connectivity app and collect all the upstream and downstream partners of these cells. IIRC, there was over 270k segments. Using Notepad++, I’ve removed all the duplicates and thinned the list down to around 170k segments. Then I’ve sorted the remaining IDs in the ascdending order.

The next step was the most tedious one. I was copying the IDs in batches of around 1000 to FW and selected manually all the cells that looked completed (FW doesn’t show the status, when there are more than 60 cells at the same time), almost completed or at least are some significant parts or cells. I’d say, over 100k of the cells were tiny fragments/twigs. During this selection I’ve also got rid of most of the big glia mergers. The result is a list of over 27k cells.

All the IDs are available here:

The reason, I’ve done it, is to find all cells of all types and identify them. If the method will be good, I might do something similar for the Lobula and Lobula Plate in the future.
Currently I’m looking for all the Lawf cells. After that, I’m planning to split the results further to 3 groups: cells, that project outside the optic lobe; cells, that projects outside the medulla; and cells, that are completely within the medulla.
If I didn’t miss any big volume, the cells, that should be in the list are: L, C, Lawf, Mi, Mt, Tm, TmY, T1 - T4, R7 & 8, Dm, Pm, Y, Olt and probably some others.

If anyone wants to join me in this searching, wants to prepare similar list for other neuropils or the other optic lobe, feel free to join :slight_smile:
If you want to go through the list, I’d suggest to work on about 500 cells at a time (unless your computer can handle more, of course).

4 Likes
#2

I am currently working on R cells (freeing them from the glial and completing them + some other cells) i have a list with one flywire link for each main type for about 300+ completed cells so far, if this would be helpful

4 Likes
#3

Wow this is amazing and a very clever way to organize identifications for a large batch of neurons. Do you mind if I share this with our lab?

1 Like
#4

Sure!
Hopefully it’ll speed up the process of identification and fixing all the incorrect cells.
I’ve found, working in batches with the same types of cells helps enormously with finding any problems with some of them.

1 Like
#5

I’ve sorted the list into 4 categories:

  • cells with synapses only in medulla,
  • cells with synapses in medulla and lamina,
  • cells with synapses in medulla and lobula and/or lobula plate,
  • cells with synapses in medulla and projections to the central brain.

Here’s a link to all four lists:

( you might want to download them all by clicking the Download ZIP button for easier manipulation).

The lists may contain segments from other categories and don’t contain all the cells from a given category. I’d say, there are above 90% of cells of each category. Many of the cells aren’t finished yet, so may contain mergers or have missing fragments. Some segments are only fragments, that I recognize as cells, that might belong to one of the categories above.

The next step for me will be to split the medulla only cells into:

  • wide-field cells (Pm and Dm like),
  • columnar cells (Mi),
  • other cells (most of them will probably be some kind a wide-field).
#6

no idea how you manage to get trough all of them or how you can search trough that list

1 Like
#7

I’m doing it in batches of 500-1000 cells and the using the Classifier addon to sort them to separate groups. Then check the groups (again, in batches) to fix some mis-classified cells. The same I’m doing right now with the medulla-only cells.

#8

well i do not think i understand it well enough too help in any way other than maybe help search for a home for some of the fragments or unfinished cells if you are making a list of that?

1 Like
#9

Yes, I’m currently in process of creating 3 groups - columnar cells, wide-fields cells and fragments, that I don’t know, where they belong to (all in the medulla).
I’d like to say, that every way of working in FW is a good way. It’s not, that I’m trying to impose my way of working on other players. I’m publishing the lists for a couple of reasons:

  • to document the process, I’m doing,
  • to share the results,
  • for others to join, if they want to.

I’m currently focusing on sorting and identifying cells, because I’m feeling more inclined to do so. But there are still many unfinished cells so the “normal” way of working in FW is also important.
But I’m happy, you want to help :slight_smile:

There are also many other ways to progress the FW. For example:

  1. Finding large wide-field cells in medulla and lobula to help identify layers in these neuropils, which would, in turn, help identifying all the Mi, Tm and TmY cells.
  2. Finding large cells in the medulla and lobulla (not necessarilly single layer thick) to cover all the neuropils. This would help in finding all cells, that synapse in these two neuropils and creating lists similar to the one in this topic but for other cells.
  3. Finding similar cells, then finding issues between them and submit their identifications (for example, many C cells still miss their extensions into the lamina).
  4. Searching for informations about types of cells, we don’t have in the Fischbach’s paper and then identifying them.
  5. Focusing on one particular type of cells and trying to find and fix them all (like @st0ck53y did with the Dm4s).
  6. There are still (at least) 14 Lawf cells, that are incomplete (link), despite me thinking, that I did them all, lol.
  7. There are: these, these, these and these cells that need finding all the other of their types, finding, what types they are (the first ones are probably ML-VPN1), completing them, if needed and submitting their identifications.
  8. Many more things to do, for sure.

I’ll probably publish the 3 lists later today or tomorrow.

2 Likes
#10

well i am working on the R (1-6) and trying to find and complete all of them in a small area, but sometimes it is good too do other stuff also to get some change

2 Likes
#11

Nice. I’m fidning working on the R cells very difficult, so it’s good to know, someone’s working on them.

1 Like
#12

Here is further division of the medulla only cells. There are four files with IDs for:

  • columnar,
  • wide-field,
  • unsure,
  • unfinished.

Of course, some of the classifications can be wrong, but most of them should be correct.
The next step for me will be either taking care of the “unsure” category or splitting the wide-field list to even smaller groups.

#13

@kk out of curiosity are you still working on that list of 27k neurons from the medulla you started with in january and in that case how far along are you? or have you found some other metod to go about it?
When using the connectivity to find cell groups, are you able to remove any cell id that you have already identified so it goes faster and not have any risk classifing the same cell two times?

1 Like
#14

I’m generating new lists almost every day, because they tend to be outdated rather quickly. Not all of course, but so many, that updating them one by one would be a trouble. To generate the new lists, I’m using the Batch Processor addon and its “Find common partners for visible” function.
To filter out all the already identified cells, I have a collection of all the cells, I have already identified and wrote myself a small tool to remove the identified cells from any list. One can also use this online tool: COMPARE LISTS - MULTIPLE LIST COMPARATOR - Venn diagram generator, free online tool to find set intersections for this purpose.
Then I’m using the Notepad++ editor. Firstly, I’m converting the list so each id is in its own line. To do it, I’m using the Replace option (Ctrl + H) in the editor, and change all “,” or ", " to “\r\n”.
After removing the duplicates, I’m sorting the list ascendingly (again, using Notepad++). Then about the first half or so of the cells are just “dust”, so I’m starting from the middle of the list and checking a few hundred cells at a time. If such sample contains only dust, then I can safely remove all the cells above in the list. Then I work gradually down the list, again with batches of few hundreds each time.
I’m using the Classifier script to collect all the cells I’m interested in, then remove the rest, refresh the FW page and paste a new batch.

1 Like