Available fly atlases

[Luigi Petrucco]

Hi natverse people, 

We started a Python project similar in scope to natverse, aiming at providing a standard set of packages for quantitative neuroanatomy using Python (https://github.com/brainglobe). 

We would provide support for a standard set of atlases, with reference brains, annotations and meshes, together with some tools for visualisation, registration, visualisation of cell skeletons, etc. We are currently all coming from either zebrafish or mouse neuroscience, and we have worked to provide support for atlases for these species. We wanted to add drosophila to the set of our standard brains, to make it more attractive to users coming from that field. 

We encountered natverse and saw that you seem to be well equipped to deal with the fly neuroanatomy, so we would ask you suggestions on the following 2 points: 

1) What are the most broadly used fly atlases that can be retrieved online and have a) a reference brain for morphing; b) complete (hierarchical) brain regions annotations and (optional, we can produce them) c) brain regions meshes? 

2) Do you know if there's already a fly community bringing forward similar efforts in python or someone who might be interested in doing so?

[Greg Jefferis]

Dear Luigi,

Many thanks for getting in touch and letting us know about your nice project and plans. I think this is certainly a growth area of neuroscience at the moment and it feels like there is scope for more across systems efforts.

On 20 Jul 2020, at 20:29, Luigi Petrucco <luigi.p...@gmail.com> wrote:

We would provide support for a standard set of atlases, with reference brains, annotations and meshes, together with some tools for visualisation, registration, visualisation of cell skeletons, etc.

Yes this is a significant part of what the natverse provides. See below for details of existing python tools with overlap with the visualisation and analysis aspect. The natverse has also recently expanded to include a lot of support for connectomics resources, which will soon enough become relevant for fish / mouse but are obviously less developed at the moment.

We are currently all coming from either zebrafish or mouse neuroscience, and we have worked to provide support for atlases for these species. We wanted to add drosophila to the set of our standard brains, to make it more attractive to users coming from that field. 

We encountered natverse and saw that you seem to be well equipped to deal with the fly neuroanatomy,

Just a note that we have made some sample packages for fish and mice; as you imply the notion of brainspaces is quite general and in fact we have been developing these things for quite some time. Some fish people have used some of our toolchains (for example the CMTK toolkit and support for CMTK registrations) but we have not seen very widescale adoption. This probably a matter of where the developer community lives but also that computational neuroanatomy probably developed faster in fly than fish although things have been quite active there for the last few years. We have seen little interaction with the mouse community, I think partly because the Allen started developing their own tools without reference to prior work in small brains (our effort was a bit less organised/visible at the time, so not surprising wrt us).

so we would ask you suggestions on the following 2 points: 

1) What are the most broadly used fly atlases that can be retrieved online and have a) a reference brain for morphing; b) complete (hierarchical) brain regions annotations and (optional, we can produce them) c) brain regions meshes? 

Take a look at 

* https://github.com/natverse/nat.flybrains

* https://elifesciences.org/articles/53350/figures#fig6 and https://elifesciences.org/articles/53350/figures#fig6s2

These days the ones that you should focus on are JFRC2  as used by http://www.virtualflybrain.org/site/stacks/index.htm and JRC2018F or JRC2018U (to which VFB will be switching). 

See also:

https://braingazer.org/

http://www.larvalbrain.org/

2) Do you know if there's already a fly community bringing forward similar efforts in python or someone who might be interested in doing so?

There are a number of relevant python projects. Philipp Schlegel in our group has produced navis

https://navis.readthedocs.io/en/latest/

starting from a more connectomics focussed library (pymaid also on github). This provides a range of computational neuroanatomy functionality. It also provides r-python interfaces to some of our key R libraries e.g. for NBLAST comparisons or transformations between brainspaces. I must say I am quite keen on this interactivity – it's great if people use their favourite language as a front end, but it feels like a bit of a waste if people spend a lot of time and effort reimplementing stuff that works well. In the other direction we have a few experimental interfaces from R to python via the reticulate package (https://rstudio.github.io/reticulate/). For example we currently use this to wrap the cloud volume python package (https://github.com/seung-lab/cloud-volume), which is used for fetching EM connectomics data (especially mesh and volume data).

All the best,

Greg.

[Luigi Petrucco]

Dear Greg,

Thank you very much for your thorough reply! There was lot of interesting content you were listing and I would like to discuss this a bit further.

Right now, the focus of our project has been much less on EM/connectomics, and much more on a light and consistent way to have available locally on a machine an "atlas" (reference, meshes, regions, etc) and easy API tools for I/O that and bunch of utilities. Considering the backgrounds of people involved, this is more thought focusing on raster data, mapping on references localisation of cells detected from light microscopy or functional imaging. We deal with meshes for visualisation of regions but so far not at all for EM. Soon enough I will need to start to work also with EM skeletons, so absolutely great to hear about navis! I will totally dig more into it and if it works as nicely as it promises I would be totally in favour of using it. I tried to play a bit with other packages supporting SWC files but was disappointed. I agree that implementing the same, exp. when talking about complicated/optimizable algorithm is not a great idea, so happy to try to interface with existing frameworks as much as possible!

As per our focus, providing fetch able atlases that people can use for registrations etc, I do agree that in many cases mouse people favour just the AllenSDK which is great but is thought to work within a full ecosystem of experimental data coming with the it; so, we were all not finding it too useful to handle other atlases. In our idea users should be able to both use atlases that we serve via a GIN repo, and also potentially distribute easily the ones coming e.g. with their publications in an easy way. We would like this to be a light pure-python API, but there's absolutely nothing python-specific about the atlases in the GIN repo, and we liked the idea of keeping it that way; they are just parsed from their native hosts (Allen, Zebrafish MPIN reference, a rat atlas, potentially fly...) munged a bit to be consistent in terms of file format, metadata, etc and then updated in the repo. If you see any space to optimise things here and potentially join forces to work with the same kind of data I would be very happy! I am trying to understand how this happens in natverse. You have the actual atlas data in repo like https://github.com/natverse/nat.flybrains right? 

I think it would be interesting to find as many contact points as possible!

Il giorno venerdì 24 luglio 2020 10:01:38 UTC+2, Greg Jefferis ha scritto:

Dear Luigi,

Many thanks for getting in touch and letting us know about your nice project and plans. I think this is certainly a growth area of neuroscience at the moment and it feels like there is scope for more across systems efforts.