-
Notifications
You must be signed in to change notification settings - Fork 8
Community Code of Conduct
When the FANC community was formed in early 2019, community members worked to manual build skeleton reconstructions of neurons through the software CATMAID. In March 2021, we provided automated but imperfect reconstructions of every neuron in FANC, at which point the community largely switched over to proofreading these automated reconstructions through the software Neuroglancer. At this point, work in CATMAID began winding down. Work in CATMAID and Neuroglancer have necessitated different codes of conduct, both of which can be found below.
Some sections of this document were derived and adapted from the FAFB tracing community guidelines written by Davi Bock, Marta Costa, Greg Jefferis, and Michael Reiser.
Our goal is to bring together interested labs to collaboratively produce and analyze the connectome of an adult female Drosophila VNC. An overarching aim is to maintain a productive and collegial community. In large part, we expect to follow the model spawned by Albert Cardona and Davi Bock and colleagues for the larval and female adult fly brain (FAFB) tracing communities.
Performing work in either CATMAID or Neuroglancer is like working with open lab notebooks. Your work is immediately visible to others, and work done by others is immediately visible to you and should be considered unpublished work of your colleagues that has been shared with you. To avoid misunderstandings, we ask that everyone in the FANC community agree to the following:
These data are part of a collaborative effort based on trust. In accepting access to these data, I acknowledge that I understand I am being given access to the unpublished work of many other individuals. I confirm I understand that using these data for any purpose without the explicit consent of the data owner is a serious breach of accepted scientific ethics. In particular, I agree I will not share information with any other individual who has not been authorized to have access to these data and who has not agreed to these terms of use.
We use Slack for communication within the community: http://fanc-reconstruction.slack.com/ Everyone actively working with FANC data must sign onto this Slack workspace regularly and check for messages relevant to them. Please ensure you have notifications configured so that you are alerted if you are tagged in a post or receive a direct message.
If you are joining the community after March 2021, you likely will not be working in CATMAID and can skip to the Neuroglancer section.
Community members may trace neurons and synapses in FANC using a CATMAID server hosted by the Lee lab. We follow a workflow pioneered by Albert Cardona and collaborators in the first instar larva TEM volume (see Ohyama, Schneider-Mizell et al. 2015 Nature and Schneider-Mizell, Gerhard et al. 2016 eLife) and in large part adopted by Davi Bock and the FAFB community (Zheng, Lauritzen et al. 2018 Cell). For tracing in FANC, we are in large part mirroring what has been done for FAFB.
- Lab heads or project leads should ensure new tracers read and agree to these guidelines.
- Contact Wei-Chung Lee to have new CATMAID accounts created.
- Lab heads or project leads should add the new person to the contact sheet and to Slack.
- The Slack channel #catmaid-edits should be used by anyone who would like to edit (merge to, correct, add/remove synapses), or continue tracing a neuron that has been worked on by another lab. A request to the person that has traced that neuron (most significant contribution), or the lab head or project lead, can be done there.
We are adopting the following guidelines from the FAFB community for tracing and derived data ownership:
- A neuron traced to identification, with the skeleton and synapses necessary for identification, having been proofread, is ‘owned’ by the tracing lab: other labs need permission to publish or do experiments based on having observed this tracing. We expect individuals will not withhold permission without good reason. We feel that individuals should be able to trace neurons that were identified as being either upstream or downstream of a neuron owned by another lab, but these efforts should be coordinated.
- A neuron that is being actively traced, but has not yet been traced to identification, is not ‘owned’. However, if a good start has been made, a lab that wants to help out needs to consult with the lab that initiated the tracing before joining in.
- If two labs’ tracing efforts collide in the same neuron, they are now de facto collaborators: they co-own the neuron, assuming that both continue to trace/proofread it to (at least) identification, or use the data for analysis.
- If a neuron is started but the lab has not been engaged with the data for 6 months or more, no such request needs to be made; however, as a courtesy, the lab continuing work on the neuron traced by another lab should notify that lab. “Engagement” is a qualitative term, but would include: actively tracing the neuron; analyzing it; or including it as part of a submitted manuscript, or a manuscript in revision.
- Even if a lab has relinquished ownership of a neuron, contributions to the tracing should still be acknowledged in any talk, preprint or publication of the data. For more information on this please see the Publications section.
- In places where these guidelines are ambiguous, it is up to the individuals involved to work something out. Good will and communication are key. In the case of any dispute, Wei is happy to offer mediation, and a consensus will be sought. In the event a resolution cannot be agreed upon, Wei-Chung Lee will make a final decision. One possible outcome is the assignment of mandatory collaborator status (up to and including publication).
We trust community members to exercise good judgment based on context, and contact the initiating or relevant lab head if in doubt. If groups are not able to abide by the good will expected of a collaborative communal effort, we may terminate a lab’s membership in the FANC reconstruction community.
- If your work was dependent on pre-publication access to the FANC dataset and our CATMAID infrastructure, it is reasonable to make Jasper Phelps and Wei-Chung Lee co-authors on publications describing the work, reflecting the effort required to make the dataset available to you via the hosted CATMAID server. If other members of the Lee lab were involved in teaching you to trace and interpret the neuropil anatomy, they should also be considered for co-authorship. Other members of the community who made significant contributions to your project should also be considered for co-authorship. Acknowledgements for lesser contributions are also appreciated. In the event of uncertainty please get in touch with Wei-Chung Lee.
- Preliminary results from CATMAID tracing may be shown in talks, at conferences, and in preprint servers such as bioRxiv. If the results shown include significant contributions from other labs, you must request permission from that lab’s PI in advance.
- For the purpose of acknowledging the tracing contribution from other labs in publications, we suggest using time invested (creation, edition and review) to identify authorship versus acknowledgement. Time invested can be specific to the relevant neuron sub-arbour, not necessarily the whole neuron.
- Analysis that has benefitted from tools created by others in the community should cite the relevant publications/packages/code, etc.
- We will require that all publications arising from work in this common tracing environment release the relevant skeletons as swc files with the publication. The connectivity information for these skeletons (synapse location only, or including partners) should also be released if it supports the published results. If not, it can be held back if there is ongoing work that depends on it, and a publication that includes this data is planned.
- Publications must conform to the spirit of the tracing guidelines given above. In case of doubt, please consult possible stakeholders. In particular, permission from data ‘owners’ must be obtained before showing images of their tracing work in any venue.
Your tracing efforts may be used to provide ground truth data for training and/or measuring the performance of automated segmentation pipelines. Metrics resulting from these analyses may be published, but your reconstructions will not be, without your permission. Similarly, your work may be visualized and rendered for purposes of making snapshots of the collective community efforts. On a regular basis, snapshots of the collective community efforts (visualizations including movies and metrics of overall reconstruction progress) will be made available to the community for use in talks, grant applications, etc. These images may be used in reviews or overviews of the community tracing effort, but are not intended for use in publications describing or analyzing a given circuit of interest.
To be clear, any use of your work to analyze circuit structure, connectomics, etc. does require your active agreement and collaboration. If you would like to opt out from having the contributions of your lab included in bulk analyses, please contact Wei.
If a lab has ceased to use the FANC CATMAID server in a substantial way, we request that the lab notify us so that all relevant accounts can be deactivated. We may also reach out proactively to wind down these accounts as well. Skeleton reconstructions of neurons can be exported as swc files and taken prior to account deactivation. The use of such files would be subject to the code of conduct described above (i.e. if others have worked on those neurons, they need to be consulted and treated as collaborators).
This section written by Jasper Phelps and Brandon Mark with approval from John Tuthill and Wei-Chung Allen Lee.
During 2020 and early 2021, we (Zetta AI and the Lee and Tuthill labs) worked to produce an automated segmentation of all the neurons in FANC. Zetta AI, in collaboration with members of the Seung lab and Allen Institute, built a Neuroglancer-based user interface for proofreading the segmentation, very similar to FlyWire but for FANC. We invited the FANC community to start proofreading the automated reconstructions in March 2021. Proofreading automated reconstructions is much faster than manually tracing neurons, and with this speed-up, community members will be able to reconstruct more neurons and larger circuits. This increases the likelihood that different labs’ interests will overlap, necessitating updated policies and more frequent communication between labs.
We expect that most labs will want to first dive into reconstructing circuits of interest in order to answer specific questions relevant to their lab’s work (e.g. the goals listed in the proposal that labs submitted when they joined this community). We encourage and support these kinds of targeted efforts. Along the way, we also encourage everyone to correct obvious segmentation errors in neurons’ backbones that they see as they work. Such corrections often take just a few seconds each, and they will benefit the whole community and the eventual effort to proofread the backbone of every neuron.
After completing their initial proofreading projects, some labs may be interested in contributing to a more comprehensive/unbiased proofreading effort where we collaboratively attempt to correct every error that affects any neuron’s backbone morphology. Such a project would allow us to produce a draft of the whole female VNC connectome and to publish this as a resource, similar to the Janelia hemibrain. Labs that think they would be interested in dedicating proofreading time toward this type of centralized and comprehensive effort, which would merit authorship on a publication reporting the draft connectome, should contact Wei. Our decision of when to organize and launch a centralized effort depends strongly on how big of a team is interested in being involved.
Groups who have already been tracing in CATMAID will likely already know where their neurons of interest are in FANC. For other groups, you will want to look around in the automated segmentation in hopes that you can find your neuron(s) of interest by eye without needing to do any proofreading. However, in order to find your neuron(s) of interest it’s likely that you will need to do some number of merge and split operations to proofread some neuronal backbones. We refer to this as the exploratory proofreading phase. Some of your exploratory proofreading will end up enabling your formal proofreading projects (see next section), but some will not, and instead they may eventually enable projects by other groups. You should not expect that the time invested in exploratory proofreading efforts will lead to authorship on another lab’s paper, even if a lab ends up doing a project that benefits from the exploratory proofreading you did. Doing exploratory proofreading is simply part of working in a collaborative reconstruction community. Keep in mind that other groups’ exploratory proofreading may benefit you in the same way that your exploratory proofreading may benefit them.
Once you identify your neuron(s) of interest, you must post on the #proofreading-projects channel of fanc-reconstruction.slack.com to describe generally why you are interested in this neuron and what kind of data you think you might want to include in a paper. Here are two examples of what we imagine these posts might look like:
I am interested in knowing the {presynaptic/postsynaptic} partners of {some type of neuron, for which you share a light microscopy image if you have one}. I have identified this type of neuron in the FANC segmentation: {neuroglancer link}. I am planning to proofread this neuron and its {presynaptic/postsynaptic} partners in order to find its strongest {upstream/downstream} partners. If anyone else has interests that overlap with this, please let me know.
I am interested in characterizing the morphologies of all the {lineage X / some other description of a cell type} neurons. I have identified some neurons of this type in the FANC segmentation: {neuroglancer link}. I am planning to proofread the backbones of all neurons of this type to see how many there are and to characterize the different morphological subtypes. If anyone else has interests that overlap with this, please let me know.
If no one else responds to indicate that they have an interest that overlaps with yours, you have preliminary ownership of the data that you intend to collect through your proofreading. You may then embark upon this formal proofreading project that your post described with the expectation that you will have priority to publish the data that your project produces. If instead another lab replies indicating that their interests overlap with the scope of your proposed project, you must discuss with them and come to an agreement about how to collaborate. Please discuss how to divide proofreading effort, and how the resulting data might either be split and included in separate papers from each lab or included in a collaborative paper. One instance of this happening and being handled well can be seen on Slack here.
Once you finish your proofreading project, that is, you have results in hand that you expect to eventually include in a publication, you must announce this on Slack to finalize your full ownership of these results. To do this, please reply to your original #proofreading-projects post with a simple message that looks something like this:
I completed the proofreading project described in the post above. We now know {all postsynaptic partners of DNz99 / the morphologies of all the T1 LinX neurons}. We are working on a manuscript about these neurons in which we plan to include these results.
Please check the box next to “Also send to #proofreading-projects” before sending your message so that everyone can see your post. In this post you do not have to announce what you found, but please reiterate what kind of data you collected, i.e. which neuron(s) are involved and whether you’re planning to report any results about their presynaptic partners, postsynaptic partners, etc.
Posts on #proofreading-projects will serve as important records of what proofreading projects are ongoing and which ones have been completed. Because they are records, do not use Slack’s “Edit message” feature on any posts in #proofreading-projects. (We will keep separate copies of these posts as a backup, but still, this rule will make life easier.) If you need to correct or add to a post, you can do so by posting a reply message to your initial post (instead of using the “Edit message” button).
We discourage individual people from simultaneously undertaking multiple formal proofreading projects. We expect that after you start a proofreading project, you either finish it or announce that you are no longer going to work on it (in which case the preliminary ownership expires). If you have a good reason for wanting to be involved in multiple proofreading projects at once, please post about this on #proofreading-projects and we can make decisions on a case-by-case basis. On the other side of the coin, you are certainly free to have multiple people involved in the same proofreading project, but try to make it clear in your Slack posts who the “project lead” is, if there is one person overseeing the effort.
At all times, you are free to proofread any neuron in the dataset, regardless of if that neuron is part of another lab’s ongoing or completed proofreading project. We have a few reasons for adopting this policy:
- If we allow labs to “lock” their neurons of interest, where no other labs are allowed to make edits to them, the number of “locked” neurons across the dataset may grow to a point that makes it a huge headache to actually do any work. We strongly wish to avoid this situation.
- We do not yet have infrastructure in place that would make it easy to know which neurons are part of another lab’s proofreading project and which aren’t. Therefore it is not logistically feasible to know when you are making an edit that would affect a neuron of interest to another lab.
- We want the community to get used to the fact that segmentations are always changing. For instance, at some point we may generate a new and improved automated segmentation of the whole dataset, and we want labs to be prepared for that possibility. Specifically, the “no locked neurons” policy encourages that labs design their analyses in ways that are robust to neurons continuing to change in slight ways over time.
- There is an easy-to-use “time travel” tool that straightforwardly allows you to see the state of the dataset at a past time. This means that labs always have the option to note down a specific time point that they want to use for analysis, and that analysis will then not be affected in any way by other users continuing to improve the reconstruction of that neuron. While this is an option, we discourage this and instead hope that labs will always make use of the most up-to-date data for their analyses.
While our policy is that there are no “forbidden edits” or “locked neurons”, if you happen to notice that you have made an edit to a neuron of interest to another lab, sending the relevant person a Slack message about your edit may be appreciated, and so we encourage you to do this. Don’t bother doing this for edits that add or remove a tiny fragment, but, for instance, if you merge a reasonably large branch that was missed, sending a message about that edit would likely be appreciated.
We do not currently have plans to support proofreading of the automated synapse predictions within Neuroglancer. See here for an argument about why manually proofreading synapses one-by-one is not believed to be a good use of time, even if we had the infrastructure for performing synapse proofreading, which we currently don't have and don't plan to work on.
Some of the text below is inspired by FlyWire’s “Fairness” principle.
Please read carefully the section above "Starting and finishing formal proofreading projects to ensure data ownership" so you are familiar with the meaning of preliminary ownership and full ownership.
-
Before including any sort of unpublished information/data from FANC in a conference presentation (poster or talk), users must have preliminary ownership of the data they will present (that is, they must have made a post on #proofreading-projects to indicate the start of their work on some set of neurons). Additionally, you must follow the last bullet point in this list.
-
Before including any sort of unpublished information/data from FANC in a scientific publication (preprint or journal submission), community members must have full ownership of the data they will present (that is, they must have made a second post on #proofreading-projects to indicate the completion of their work on some set of neurons). Additionally, you must follow the last bullet point in this list.
-
Before including any sort of unpublished information/data from FANC in a conference presentation (poster or talk) or a scientific publication (preprint or journal submission), community members must obtain agreement from every principal investigator whose lab undertook a formal proofreading project that contributed to those reconstructions. It is every community member's responsibility to read the posts on #proofreading-projects to be aware of which labs they must consult. Labs may agree to share their work freely, may request acknowledgement or co-authorship, or may request that the data not be shown or used for the time being. Negotiations between principal investigators may be necessary to reach an agreement, but the bottom line is that an agreement must be negotiated. If at all possible, be generous with sharing data and be generous in including others as authors when they share data with you. Realize that working in a collaborative community means being more liberal with sharing data and including others as authors than you might be used to from other contexts.
To be explicit, full ownership of data (which results from completing a proofreading project) does not mean you get absolute unilateral control over the data – however, we expect that no one will make aggressive requests to use data that another lab has full ownership of. If a request comes up while you have only preliminary ownership of some data (that is, your proofreading project is in progress), we encourage you to view this as an opportunity to collaborate toward accomplishing a proofreading project more quickly. However, in all cases, if someone makes a request that you object to, you have every right to decline the request and attempt to negotiate an alternative agreement.
When you are a few weeks away from having a finished manuscript, please head over to Data release checklist and take care of the steps there at least a week before you post your preprint / submit your paper.
If an agreement cannot be reached, a committee composed of Wei-Chung Lee, John Tuthill, and a third member of the community (to be determined on a case-by-case basis depending on the individuals involved) will arbitrate the dispute. (In disputes involving the Lee or Tuthill labs, an additional community member will take their place on the arbitration committee.) The following are three of the main principles that will to be taken into consideration during arbitration:
- The guidelines in the sections above clearly state our expectation that you communicate early and communicate often about the proofreading projects you’re working on within the EM dataset. If a conflict arises between two groups and it is clear that more open communication from one group could have helped avoid the conflict, that will weigh heavily on the committee’s decision.
- Work done to determine a neuron’s identity (e.g. by comparing with light microscopy data), often done during the exploratory proofreading phase and posted on Slack when a proofreading project is started, is seen as an important contribution and will weigh into the committee’s decision.
- The more time that a lab invested in a proofreading project, the more the committee will be willing to protect the rights of the lab that performed the proofreading. For example, a lab that undertakes a formal proofreading project that requires only a few days of work is unlikely to be tolerated if they insist that other labs cannot use the resulting data in any way until it is published. However, a lab that undertakes a proofreading effort that takes six months of work may be seen as justified in asking another lab to delay their use of the data for a significant period of time, potentially until the first lab gets a paper accepted.
We view arbitration as a last resort. We hope to never need it, but we understand it may be the most logical way to resolve some situations. In the unlikely case that a single lab is repeatedly involved in disputes that require arbitration, we reserve the right to revoke access to the automated segmentation for specific labs.
If your project benefited from early access to the FANC automated segmentation (before our preprint was posted to bioRxiv in December 2022), we ask that you add Jasper S Phelps, John C Tuthill, and Wei-Chung Allen Lee to either your author list or your acknowledgements section. We leave this decision to each lab to make based on their standards for authorship and the extent of the collaboration.
- For projects already well underway in CATMAID, there are two main options:
- You are welcome to complete your projects by continuing to work in CATMAID. It is possible to export automatically segmented neurons as skeletons and import those into CATMAID to manually join them with your existing tracing, but we have not done this yet ourselves. (Greg Jefferis’s group has done this some in FAFB and may be able to provide code and advice if you’re interested in doing this.) Note that if you use the automated segmentation in this way, you will then be subject to the publication embargo, whereas if you just complete your project in CATMAID (while looking at the automated segmentation to guide your tracing efforts if you find that useful), you will not be subject to it.
- An alternative plan would be to start a segmentation proofreading project targeting the neurons you have been tracing in CATMAID. Groups wishing to do this will certainly get priority on undertaking such projects. (That is, if you’ve been spending lots of effort tracing some neurons in CATMAID, we will not tolerate other groups attempting to work on the automated segmentation version of your neurons before you’re able to complete your project.)
- We highly recommend that groups not start new manual tracing efforts using CATMAID.
- While it is possible to export neurons from the automated segmentation as skeletons and then import them into CATMAID where various skeletons can then be connected together and extended via manual tracing, we strongly discourage this approach, except for the situation where projects are already well underway in CATMAID (detailed above). We believe everyone will be better off if we agree to work collaboratively on proofreading the automated reconstructions, instead of importing fragments into CATMAID and doing merges and cuts there.
- Once a neuron has been proofread essentially to completion in neuroglancer and is nearing ready for publication, we can then help import it into CATMAID where final edits can be made if desired (e.g. correcting a few synapses or twigs). We require that you make skeletonized versions of proofread neurons available on Virtual Fly Brain as part of your publication, as has been standard for papers that include skeleton reconstructions from FAFB. We can help you with this process.
We expect to add new sections or make modifications to this document as interests, infrastructure, and technology related to FANC reconstruction evolve. We invite the community’s input about when additions and/or modifications to these policies are needed. Please feel free to use the #neuron-ownership-discussion channel for ideas that you wish to get community feedback on, or email Wei directly. Ultimately, we reserve the right to make final decisions about proofreading and data ownership guidelines, but our main goal is to enable scientific progress by keeping the community harmonious, productive, and free of conflict.