Where communities thrive

Join over 1.5M+ people
Join over 100K+ communities
Free without limits
Create your own community

Explore more communities

openworm/open-worm-analysis-toolbox

A testing pipeline that allows us to run a behavioural phenotyping of our virtual worm running the same test statistics the Schafer Lab used on their worm data.

People

Repo info

Activity

Eviatar

@Eviatar

I'll let Andre weigh in. He may have strong opinions having faced this problem multiple times.

Chee Wai Lee

@cheelee

I'll need to follow up (as best as I can given a lack of access to published literature) with the cited publications, but has there been any other comprehensive large-scale data analysis work on worm movement like in Andre's paper http://biorxiv.org/content/early/2015/04/08/017707 ? Naively it feels to me that an often repeated sequence of shapes-over-time can be used as a context-neutral method to further characterize context-sensitive movement behaviors such as the Omega (full disclosure: I have no clue what is meant by an omega turn.)

The reason I asked is because in my previous field of application performance analysis, a very similar problem plagues the research community where characterizing the (high level) "behavior" of various scientific code kernels is concerned.

balazs1987

@balazs1987

hey Ev,

thank you for your comments

generally we are in agreement

regarding the difficulties in defining what is an omega turn

In my last paper (http://rsif.royalsocietypublishing.org/content/12/113/20150899) I have argued that behaviour is not a set of discrete states, but rather a continuous spectrum. When presenting my results people often assigned the observed continuity as an artefact of my method rather than as a feature of behaviour. I have been criticized that other omega turn detection algorithms or experts could pick out omegas unambiguously.

I wanted to put this claim to the test and this survey is a part of this effort. I have already compared 4 omega turn detection algorithms (Laurent 2015, Yemini 2013, Huang 2006, Salvador 2014) and they often have a disagreement for over 50% of the events!!! I am just collecting the expert annotation now, but based on what i have seen so far that is going to be pretty diverse as well.

By highlighting that neither expert annotation nor the algorithms are consistent with each other I hope to emphasize that behavioural annotation is actually inconsistent in the literature. Furthermore I think that if we can not agree on an omega detection algorithm and expert opinion is diverse as well, then that would be a strong argument to consider the 'behaviour is a continuous spectrum' framework as an alternative to the currently dominant 'behaviour is a set of discrete states' paradigm.

one more thing: if you could invite your lab members to the surve I would be very grateful

I had many responses from PI's, but for some reason it seems that people have not forwarded the invitation to grad students in their labs

Stephen Larson

@slarson

(cc: @Eviatar )

Michael Currie

@MichaelCurrie

Hey guys we are meeting now on Google Hangouts. Let me know if you need the link

Michael Currie

@MichaelCurrie

Agenda: https://docs.google.com/document/d/1ueJvJ1vHfiyHiKO5gHNt8x3F0qhFUpDZRZP7t2rt60o/edit

Michael Currie

@MichaelCurrie

Thanks for attending the meeting on 10 June. I've set the next meeting to be Friday 15 July at 17:00 UTC and sent out a calendar invitation. Andre, Jim, Rex, let me know if this date doesn't work and I can move it.

Chee Wai Lee

@cheelee

@/all Hey guys, this came up during a hangout with Dr. Barry Aprison from the University of Chicago and the Education Committee - do we have any movement datasets where worms are subject to various forms of stress, in particular Osmotic Stress scenarios?

Andre Brown

@aexbrown

@MichaelCurrie I should be able to make it on Friday, but I'll be at a meeting so it will depend on exactly what's going on that afternoon and whether I feel I need to attend.

Michael Currie

@MichaelCurrie

@aexbrown ok no problem. meeting starts in 40 minutes, it may end up being a short one, we can just quickly touch base.

@cheelee I'm not aware of any such datasets but someone else may know of one.

Michael Currie

@MichaelCurrie

Hi Team, we just had our meeting (Jim and Rex and myself). Here are the minutes: https://docs.google.com/document/d/1qZII5HoKUL2_HrhkFl_t8zblTW9CsCgBCDI3eyHgbIs/edit. See you at the next meeting, on the 12th of August.

Michael Currie

@MichaelCurrie

(I set the next meeting 2 hours earlier than usual to better accommodate our Thai, Singaporean, and British friends..)

Michael Currie

@MichaelCurrie

Hi Team, we just had our meeting (Jim, Rex, Chee Wai, and myself). Here are the minutes. See you at the next meeting, on the 15th of September.

(note that the next meeting is on a Thursday - not our usual Friday - to accommodate everyone's schedules)

Michael Currie

@MichaelCurrie

@ver228, @aexbrown: FYI we are hoping to get your opinions if possible on any final revisions to the tracker-commons format, at our 15 September meeting, ahead of further publicizing the format. Hopefully WCON will be more "battle-tested" by then since @ver288 will have done plenty of file conversions to that format by then. Thanks!

Chee Wai Lee

@cheelee

I'm getting this weird thing where Travis-CI is reporting testing errors since a few check-ins ago for OWAT, but quite inexplicably github is now allowing me to merge my pull request (of metadata, which shouldn't affect the testing framework at all.) Anyone has any clue what's going on here?

Michael Currie

@MichaelCurrie

Hi Team, we just had out meeting (Jim, Rex, Chee Wai, André, Avelino, and myself). Here are the minutes. See you at the next meeting, on Thursday 13 October.

Michael Currie

@MichaelCurrie

Hi Folks, no one was able to make the 13 October meeting, so I have rescheduled it to 10 November at the same time. You should have a calendar update in your email inbox. See you then!

Michael Currie

@MichaelCurrie

Hello, because of a couple of conflicts from people, I have rescheduled the 10 November to 24 November. You should have the new invitation in your email inboxes. Thanks.

Chee Wai Lee

@cheelee

@MichaelCurrie Saw that. Thanks!

Stephen Larson

@slarson

Hi guys. Was just talking with @a-palyanov and @skhayrulin today. We now need to be calculating wavelength and amplitude of worm swimming and crawling from Sibernetic for the next paper. @a-palyanov had been estimating it from images but now he’s needing to automate this process, and so I thought we should get back to figuring out how to automate this with the software developed here. What is the best bet to get started with this, do you guys think? cc: @JimHokanson @MichaelCurrie

Stephen Larson

@slarson

There is this as a potential start: https://github.com/pgleeson/sibernetic/blob/development/src/plot_positions.py

Screen Shot 2016-11-11 at 8.25.48 AM.png

This is the output that comes from this

@a-palyanov would this help?

a-palyanov

@a-palyanov

@slarson , thanks, but I'm already getting C. elegans body posture profiles directly from Sibernetic after a small modification of its source code.

Stephen Larson

@slarson

cc: @pgleeson

Stephen Larson

@slarson

@a-palyanov and @pgleeson -- @aexbrown is telling me if we can output in the WCON format, then there are a lot of analytics that can be done easily / for free. Can anyone in this channel point to a quick tutorial on how to encode in WCON for these guys? Thanks!

Padraig Gleeson

@pgleeson

Looks good @a-palyanov. How can I get that out of Sibernetic?

@MichaelCurrie @aexbrown does WCON and compatible tools require points along both edges of the worm or is a midline sufficient?

Michael Currie

@MichaelCurrie

@pgleeson midline is sufficient

See the specification here: https://github.com/openworm/tracker-commons/blob/master/WCON_format.md#data

"The body position of an animal is given by x and y. These are typically arrayed and represent points along the midline of the body. "

ver228

@ver228

Hey, I was talking with @aexbrown about adding the dorsal-ventral orientation to the eigenprojections. This is equivalent to multiply by -1 one of the orientations (either dorsal or ventral). For what I understood this is not in the code, but please correct me if I am wrong. I was wondering if the versal-dorsal orientation information is already being used. I think some of the features (bends) should have already incorporated it.

Michael Currie

@MichaelCurrie

From what I understand this is not in the code, as you say. Perhaps @JimHokanson has something else to say though.

Michael Currie

@MichaelCurrie

Hello, the team just had our monthly meeting. Rex, Chee Wai, André, Jim and myself spoke for an hour about the past month's developments. Notes are here: https://docs.google.com/document/d/1hFumS02rujQa_YGC1fBfUV8Dg4WrX057f4NnahEY4rw/edit. The next meeting is on 22 December and is in the OpenWorm calendar. Thanks

Eviatar

@Eviatar

@ver228

Perhaps I'm misunderstanding but dorsal/ventral have no meaning for eigenworms. I signed features wherein the feature itself had an orientation relative to the dorsal/ventral positions of the worm (e.g., a bend who's convexity was oriented towards or away from the ventral side of the worm -- since worms crawl on their left/right side).

The eigenworms are a projection into a new linear coordinate system that permits dimensional reduction without losing much data.

I'm not sure what dorsal or ventral would mean in this context.

If you change the sign of the eigenworms you will change the loadings in the projection & thereby lose the data necessary to reconstruct the angles of bending along the worm. But the eigenworms are not these angles.

Eviatar

@Eviatar

Generally, I think the eigenworms are a confusing construct & perhaps should be hidden well inside the WCON so that only engineers who really need this data would bother extracting it. I frequently get questions about significance in the eigenworm features & have to explain that they do not easily map to human relatable concepts of phenotype. They were specifically chosen by Andre as a non-biased means of extracting phenotype. The other features we provided are highly correlated with the eigenworms & present a better means of describing phenotypic differences in a human-centric manner.