DevoWorm Lab Meetings, 14 Sep 2020

Previous Meeting Next Meeting

⏯

youtube image

►

From YouTube: DevoWorm (2020, Meeting 33): Community-building and Gene Regulatory Networks, Shape, and Form

Description

Attendees: Bradly Alicea, Mayukh Deb, Mainak Deb, Ujjwal Singh, and Jesse Parent.

A

A

Hello, how are you.

B

Hey uh good morning.

C

All my vacationing is officially over the first year.

C

D

C

From from here,.

A

Well, that's good yeah, it's good to take some time off and or you.

D

A

Doing stuff during your uh vacation, but it's always good to have like.

D

A

That's different.

C

Yeah, um definitely, you know, let's just attach this mind, shift around a little bit. uh So that's that's good and I did a little bit of stuff, but then, as the week that.

B

Last week, progressed.

C

That's some other things, but more formally doing the conditions like.

B

C

What it's just it's not gonna, happen right now, um not at least not in the full form, and I should just wait until now. It was just you know, it was just a little bit too much at once, um but all that still in the works, and I still want to figure out specifically what to do.

C

My big question is how you know: I'm still, I'm still trying.

B

To figure out what specific.

E

C

For both of the groups um for the fall and kind of through the rest of the year, I think that may be my my overall goal for next week, but.

A

I don't know I don't know if anyone else is coming to the meeting, but um I was trying to rise some people here today, but uh I think the we'll just do a recording and see where uh I have a couple things to cover and then, if people see it on youtube, then we can discuss it further, uh but I can go back over. I think you missed the talk on the project, so I can go over those really quickly as well.

A

So all right, if anyone comes in, they can join us in progress, otherwise I'll be recording this and I'll be on youtube later. So the first thing I wanted to talk about was: I have some slides on community building in and jesse might find that interesting.

A

So why don't we look at that real quick? So this is the community building. I have some thoughts on this um I'm going to share my screen, so I can so you can see what I'm.

A

A

A

You see my screen.

C

Yeah all right.

A

So this is something I put together. uh This isn't like definitive and I'll probably come back to some of these things, so some thoughts on community building. So how do you build a thriving community of contributors and I'm actually currently facing that dilemma right now? In another context, uh so you know there are a lot of things you can do.

A

um You can concentrate on getting people in the github and committing things to github and yeah making uh you know contributions there, so it could range from like contributing to code to some sort of like what we've been doing in the uh saturday morning. Group which is to you know, work on a lot of different things like some of the uh frontier maps and things like that.

A

um So that's one way. Another way is to have regular events, so this is an open source friday, which it's a common theme in community building. You have a a session sometime during the week.

A

People can come and it's like a hackathon, but they can come and talk to the leaders or collaborate on things with other people or whatever and that's a regular time so people you know people know when to to show up to ask questions, then you can have an actual hackathon, and so this is something that we haven't done too much in open worm lately, but um I'm planning one right now for in a different context and I'll make a new I'll break that news when it's what happens, but this is something that I just actually just attended a hackathon a couple weeks ago uh through e-life innovation, and they did this all virtually so it's like you know you have like a chat room.

A

You have like a video interaction. You have a slack channel. You have to coordinate all those things then. Finally, you have to have some sort of promotion system here. So this is a user contributor committer. You can see here that over time someone goes from being a user to a committer, so they're actually gaining influence and seniority in the community over time by these contra contributions that they're making and so to put success in perspective. Keep these two graphs in mind.

A

You have the lifespan of a human which is uh broken down here in months on the right and then on the left. You have this cosmic calendar, which shows like where humans are in the scheme of the of the cosmos and our entire existence as modern humans has been in the final minute of the cosmic calendar. So we're not maybe that significant, but we're also, we have a lot of time and it just depends on how we manage that time.

A

So I mean I've worked on open mentorship models. That's what I do in in diva worm. That's what I do in the other group. uh This is inspired by distributed open source project development, so small contributions can move along larger contributions. So we pick some sort of like introductory task. That's really simple that people can do quickly and then you know maybe a short talk or some skill building task, and then we take those contributions. We say: you've gotten.

B

A

Congratulations, we give them some sort of ownership of something, and then you integrate people into the larger project and you have to understand when the interactions are difficult.

A

So if people tune out or if people are not doing well and they give up on it, then you have to understand where the point of failure is- and I mean I fail at that all the time and it's just a matter sometimes of timing or of like you know, people's schedules, but it's important to be able to at least understand when that's going to happen, so you can drive projects forward and then the mentor assigns the broad outlines of the task.

A

Ideally, and the mentee defines the details of the task, so people come into the community, they often pick up little details of a project, that's ongoing, and that gives you you know a way to build a larger project around their contributions, and so this is our lab. This comes from our lab contribution philosophy. This is like it. This basic idea broken down, so you have this early contribution.

A

You do the work, you fail and succeed in different ways. You redefine your contribution and then you do things like you might author, some paper or code, you document things and in wikis or in a presentation, and then you release your collaboration to the world. Ideally it's you know iterative. So you have this feedback loop here.

A

So we we get feedback and group meetings from other group members and things like that.

A

We can also leverage our educational stack, which is the I think I showed this in a previous meeting where you have uh communication channels and annotation, and you know.

B

A

Like video of someone giving a lecture participating interactively and then getting feedback on your assignments and things like that, um but then there are four other ways you can do this. You afford basic tools. You have. There are more tools, of course, but I'm going to talk about four here.

A

You have github where you can recruit people to make pull requests, address and plan issues and collaboratively write papers. So you also have twitter, which is to recruit people to events. Now you don't have to do this through twitter. You can do this through any social media, but with the neuromatch in the neuromatch case, which is something that we did last summer.

A

I found twitter to be very useful, and so this is uh jesse is uh tweeting on behalf of the orthogonal lab and we were organizing a study session for neuromatch students, and this was a mechanism for both them to benefit from uh having a group of people around using our infrastructure and to get maybe to get people into the lab, get them interested in things surrounding neuromatch.

A

So it can be a you know. You can use an event like neural match to be sort of uh symbiotic. In that way, you can also recruit people to a slack space to chat or collaborate, and so we have maybe a new slack for divo learn. um I have this empty slack, that's ready to go. I might actually use this as maybe a broader slack for like some of the other collaborative efforts or hackathon efforts that I've been organizing and jesse.

A

You might be interested in using this slack for some of the things like the cognition uh group and I'm not really sure how to pull this off, because I hate slack proliferation. It's like you know. If you have one channel, it's like easy to stick to that. One channel I mean you know. If you install the slack app, you can monitor all the channels that you want it's a little time consuming.

A

But to have everything in one place is the best policy I think, but sometimes you want to have a different channel or different team, for you know uh sort of vertical integration if you will- and so this is open, I right now- I'm just thinking about it using it for diva learn, but I'm thinking maybe making it a little broader, uh maybe like something like hackathon and community. You know something that's broader, to incorporate more people from different places.

C

I'm actually in.

A

The same thing, I don't want to.

C

A

Your stream too much but.

D

C

B

C

Like you know,.

C

For different groups, because well but.

D

In my case, I'm thinking.

C

There's, I think, for me it's about a space where the people the people focused in.

B

My case, I think.

C

B

Of it, because.

C

D

Different people in retirement and different products that are artificially related all over.

B

The lab- and it's like how do I give.

C

That community a space- and it's not.

B

Like not too much, but at the same time, what does that mean.

B

C

Like no, it's gonna.

B

Happen but it's sort of like.

C

How do you get the key, how.

B

Do you find you like my question right now is how do I.

C

Give a space for the community people.

B

C

B

Are going to be active and engaging.

C

And how kind of headquarters be able to do it? Because.

B

I feel like that's that's what I need.

C

B

I feel like it's very different.

C

B

C

What the critical focus of what you want for the community or the project, it's it's kind.

B

C

There's a lot of moving parts yeah.

A

Yeah and slack is only one option, but I like slack in terms of like having the team like building a team within slack, so you have channels and a lot of people use it. um You know- and it's like I said, uh tool. Proliferation in general was bad because you're checking all these different channels and monitoring them, and then the people use them. You know it's like you have enough fragmentation uh potentially to have you know no real communication, so you have so many channels. That's the risk.

B

A

So this is a way I don't know how to brand things exactly. Maybe you want to leave everything in uh the open arm slack, but you know when you're trying to scale up a contributor community, um then the.

B

Real question is.

A

Where do you want to put all those people slide them into different tools and get people? uh You know get some interactions going. That's that's the challenge so um and then we can broadcast events on youtube. So this is asynchronous updating. We do this with this group with the saturday morning, neurosim group, and it's just a way to like get people to see it if they can't make the meeting we deal with people in a lot of different time zones.

A

So one way to sort of facilitate uh synchrony is to give them opportunities to see view things offline, or you know, to have like slack of course again is asynchronous as well, where you can like just put things in a channel or you can do things synchronously, um but not everyone can do everything at the same time now, even in like you know, there might be a common time between the europe and north america, where you have like.

B

A

Morning afternoon, uh setting where people can make it during waking hours, but people are doing different things at those times, so it's hard to synchronize um and then finally, I wanted to mention open source licensing. So this is, uh I think we didn't talk about this during gsoc, and we probably should talk about this at some point uh to point out that there are different ways to do licensing for open source projects.

A

You have creative commons, which is this licensing is misspelled there, but anyways uh you have the creative commons licenses, which are what they call copy left which have this uh they're different. You can have different types of creative commons licenses for different needs, but they basically open up a project and say we're starting from the premise that this is open to share and but we're going to put restrictions on it.

A

So you start with cc0, which is this basic uh license to something really specific, like ccby nc and so that's uh creative commons by attribution, which means you have to cite things in non-commercial. So if you ever use this material, you can't use it for commercial gain and you have to attribute it. That's the rule and then it's it's pretty open.

A

Otherwise, but it's um it's legally binding, just like any other like copyright, would be um you have divorm is operates on a ccby license, so I have actually I'll show a license in a minute from the repo and then software packages in general are often mit open license.

A

So like if you're doing like looking at something like um one of the open source software packages, uh you know they usually use mit open license and it's a different. It's not based on copy off so much, but it's specific to software design.

A

So this is the. Let me see if I can find here it is. We have a repo in diva worm, licensing drm, and we have this ccbysa license that okay, so we're at bysa, uh and then this is version 4.0. So this is the uh as of 2018. This was the latest version of the license and it may have updated. We might update this uh soon. So this is what it looks like you just put this boilerplate into your repository.

A

You can find these online this boilerplate and then you can put it up, and everyone knows what your license is and what your terms of contributing are as well perfect timing.

F

A

Okay, that's fine yeah! We've got some things here on the on the plate to go through, so I just talked about community building and uh open source licenses um for contributing to projects. So I I can. I have the slides available here and I can put them up on uh I'll put this recording on youtube. You can watch it after I might be relevant to like diva learn stuff.

A

um So that's that's the community building stuff, um my hook isn't here, but we were going to talk about. He sent a new draft of his journal of open source science or open source software paper that we're working on so in case mayoq wants I'll, give him more feedback on um I'll, give him more feedback on slack a messaging. I'm! Okay!

A

um Maybe we'll hit that in a little bit here. Why don't we go to the task board for uh group meetings, see if he shows up we'll talk about it, then? um So again we have the group meetings task board and this way we haven't revisited this in a while- and I was like pruning it last night going through some of the ideas on here um and there are a lot of good ideas.

A

uh Some of them are sort of. I can't remember what they are, but some of them are ongoing. So there's, uh okay, present collective pattern. Generators id, I think, that's.

B

Distinct from the paper.

A

So we won't oh, there is so maybe we'll go. I I.

A

Maybe we'll go to the paper now: okay, so we'll actually I'm presenting so we'll come back to the board in a minute. um So this is the paper. The journal of open source software paper that my oak and I have been working on um so journal of open source software is uh it's like totally run out of a github repository, so they you mock the paper up in latex and you submit it to the um journal.

A

You put it in a repo in your in your github repository and then you send it to them and they evaluate it, and then they put their stamp of approval on it and it's published in their journal.

A

So the idea is very simple: you just have some software that you've built it's open source and you just write an uh description of it and how it might be useful, and I think we covered a little bit of what those papers look like in another meeting, but this uh usual- or I mean uh mayak, has uh spent some time on this.

A

I was meaning to come back to this as well, but I didn't you know I hadn't gotten a chance to yet. So. Thank you for driving this forward. My hook, um I can give more detailed feedback later, but right now, let's see what we have.

A

So we have a summary which is extracting metadata from my microscopic videos images one of the key steps in the process, so this kind of like describes it, um and this is basically it we'll, probably rewrite it a bit to maybe be a little bit more exciting or like so diva.

A

We talk, mentioned evil, learn here, uh then we have the statement of need, so diva learn 0.2.0, and so we have this package, and one thing here I can see is that we want to be clear about differentiating the diva learn platform here from the diva learn software, and I think we I don't know if we do that or not, but I can go through and make sure it's as clear as it can be, um and then, where are we all right? Then we have this third figure, which is a diagram of okay.

A

So this is a diagram of the sort of the difference. Okay. So this is the github source. This is the package. This is the lineage population model, egg, embryo segmenter, embryo generator model model files, and then it all fits together.

A

This is the user environment, so the the pip install package is the sort of the gui for this, and then we have this part of the user environment where they're kind of putting the package together.

B

A

My neck love my neck, um so yes, the diva learn has been built to be very data science friendly and has to be highly compatible with libraries like numpy and pandas, as divalearn grows, bigger with more tools and deep learning models. The combination of beginner friendliness and support for data science will make enable exciting scientific explorations.

A

That's good, um yeah I'll go through the paper and I'll sort of uh make uh suggestions for changes, I'll, maybe massage the language, and then uh I think we're gonna need some references, and I can find some references for this and it's a little hard because you're doing like software and it's like what kind of references would I need for a software package, but I think there are some that we'll use from.

A

I think there were some I was using in the when I was writing up the description of the g sock project, so I'll try to track those down and put them into this incorporate them into this draft as well. It's actually did. uh I can't remember if it was the the main uh well. I gave I put one description up on like incf's website and then I had one that was internal and I think that one is the one with all the references in it.

A

So we'll we'll bring we'll revisit that and I'll put in the references and we'll have about it'll come out, maybe to about two to two and a half pages.

A

It should be a nice paper and that's probably about the length of what they're looking for so and then we'll have to reformat it into their format as well, and that's, I think, latex, where you uh have this, you compile the paper um from some source code and you know a compile it can compile in your repo. I guess yeah we'll send move the paper to markdown. Actually, I think the way that they want the well we can. You can move it to markdown, but we get also to generate this version.

A

That's a tech version or latex where you have uh like you have to automate the formatting and but they have a specific format that they use.

A

So I have to go find that template and I sometimes I do a lot of that work in um overleaf if you've ever used overleaf it's a platform on the web, so you you know that if you use overleaf, you can actually write the latex code or the tech code in the left, pane of the window and then the right pane it renders the yeah it renders the document in real time.

A

So you just compile it and it'll. Show you how it's formatting it and yeah. Oh yeah, use it make your cv yeah, I don't. I don't have a paid version so, but I think it's still. Everything still works the free version. So it's that'll work.

A

That's usual might be a template available on overleaf, probably yeah. They they have, they have they usually make it available. Obviously they want people to use it. So I'll go look for it and we'll we'll get this going. I uh yeah it's just gonna, take a little bit of time on my part too. um So this board. We have a bunch of tasks, we have finished tasks and um a lot of these. This is all the stuff we've done this summer and early fall.

A

We've done a lot of stuff with uh making videos of nut logo and presenting on things and doing things uh presenting on some origin of life, stuff, uh g-sucks student presentations, not to mention the gsec projects and then copy cell 3d, which we kind of visited. But we, you know we're going to put that in finished, and then we have our in progress things.

A

So we have things like tutorials for youtube and I think the g suck final presentation, so it should be in here um the diva bibliography with endnote I'm working on richard gordon with that and we've got some. You know we've kind of worked out a system for putting things in. You know an endnote uh account, so you know why do we want to use something like endnote?

A

Well, you can take references and put them into these bibliography software um platforms, and it allows you to do things like sort them and insert them into papers, and things like that, and it's very easy to organize references, especially if you use them over and over again and so we're trying to assemble different bibliographies on different topics, uh not to mention a lot of the papers that we deal with here were the papers that we published, and so those are all going to be available in a bibliography form.

A

What else uh we're still working on the periodicity paper? So I was telling jesse earlier that we have a lot of outstanding tasks and we want to revisit those as much as possible so that we can drive them forward.

A

The axolotl embryo, animations and segmentation, so I did actually I don't know I talked to usual a little bit about it in the slack. I don't know if he was uh had any updates on that.

E

Oh, I've started working on that but, like I am right now creating models and like uh we have to uh make them compatible with their browser. As.

B

E

Susan, as I said so, I am working on an electricity.

A

Yeah, okay, that sounds good. I'm just just checking uh uh you know to see where it so it's in very much in progress. You can leave that there um so yeah. This is the periodicity in the embryo paper and jesse just asked: what's that, oh, the what papers in dw were available so.

B

A

Actually, the um I had a, I think I gave a talk last week on the open things and open projects for fall, so I think everyone but jesse saw this, and so this is in.

A

I can send the link to this to jesse afterwards, but I have a list of things that we have, that we need to kind of follow up on and the um the periodicity paper is. I think, right here, one of these.

A

I think this one, no this one here so this is uh like I was explaining last week uh and we have a repository for this. This is a paper where we're looking at different embryos, so this is very early zebra fish, embryo and we're trying to.

A

We have cell tracking data for zebrafish and for c elegans and we're looking at how cell division happens or cell differentiation over time, and so with the tracking data. We have data, for you know minute by minute data on cell tracking, and it's not perfect, but it gives you this idea of when the cells emerge in development, and so, as you can see, sometimes you have some very interesting patterns over time.

A

Different pulses of cell division- and you can do things like you- know, mine them further- the interval between these peaks and get a distribution of that, and so think that's what we want to kind of do with that paper. Whenever we can understand, you know we have cell division going on, so uh cells are being born, cells are being differentiated, but to what end? What exactly is the pattern?

A

You know? Are there pulses of cells or are there you know? Does it just happen, sort of uh in a way that there's no real pattern? So you can see that, like very early on here in this time, series that there are these very tight pulses of cell division and then later it's a lot more diffuse, and so why is that? Well, we don't really know, but we can look at the patterns and kind of make some uh educated guesses uh c elegans exhibit similar patterns of cell division.

A

So if we can, um I think, that's basically the idea- and I have some. uh I have a a document on github which I can send you a link to, and also a sort of a list of things that people were kind of talking about doing so we'll do well I'll, send that off to jesse um or put it in the slack channel, which might make more sense um on on open worm.

A

So we have the periodicity paper. We have the complexity measures which is a more general uh project which involves a number of different things and the complexity measures, so that was this lagrangian embryo is part of that and there are other things as well, uh not the psychophysics paper so much but the lagrangian embryo.

A

um So that fits into complexity measures. Then we of course, have the bacillary neuronal cognition paper and then on hold. We have things I heard from krishna katyal earlier this week. He said he had been sick with coronavirus, actually so best of luck to him and recovering because he's better now, I think, but he's still kind of sluggish. So he you know is he's been. These things are on hold, um create narrow, match times of diva worm, which is something we never did, which we had the summer school neuro match.

A

We might think about creating tie-ins at diva worm. There's a conference called super neuro match that's coming up in october at the end of october, and we might consider uh tying things more specific, more explicitly with diva worm. Then too, because uh you know we might make like a conference submission and then you know make some sort of at that same time, publicize some of the stuff that we've done in diva worm, that's neuro related, which is a lot of it. um So it's definitely something to watch.

A

I don't know how to put a flag on this, but that may be something that's time specific, like you know, maybe later in october,.

A

So then we have these embryo visualization tasks. uh The docker package, which we haven't, talked about in a long time and then the embryo model, one blender, which was something was working on.

A

So those things we can revisit later.

E

Yes, like this, one has been assumed yet.

A

Yeah yeah, it's uh yeah, the the embryo visualization, is still like. Don't really know what that's going to look like. That's.

B

A

The embryo model is, it wonders, tricky yeah. We want to make sure that we get that right. I guess um and then the bottle montaging that's sort of this task here, so 25 and 5 are linked.

A

So uh then this model is for biological living networks. I'm not really sure what that is, but maybe we'll put that in to do instead of hold, and then we have some other ones recruit people as diva learn contributors.

A

So I'm going to put that on hold for now. But what that is is that we have diva learn now we want to recruit people to uh to participate in it to contribute- and I just uh I'll post these slides after the meeting on um that I did on community building, and so hopefully people can look at that and think about it a little bit and you know we can use different communication channels for building a community.

A

I was in contact with someone this weekend who responded to a task they said. Can I contribute to the data.

B

A

Tutorials- and I said yes and so they're going to contribute one, I haven't seen a pull request yet, but um so that's that's something we can do um this thing by jesse, that's open. um He was gonna review. This paper, skill-free biology, integrating evolutionary developmental thinking.

C

A

Okay um also we're doing this stuff with modeling neural organoids, but that's that's not even wasn't, even in the mean diva worm group, but uh something that we've I've talked about with a couple people outside of the group and uh I'd be something people are interested in. If you let me know, I can tell you where we've been on that um and I was gonna, give a lecture on an artificial embryo.

A

I might do that. This fall uh compositionality, that's an interesting topic, but I'm not gonna talk about that for a while, as computer tomography via cloud computing idea. This was something that richard gordon proposed a long time ago.

A

I was about looking at ct, imaging or ct images and using some techniques uh kind of bringing that closer to sort of what was going on with the machine learning stuff. So I have, if you're interested in that I have notes on that. We can revisit that then. Finally, I was going to give a lecture on pca, umap and t-sneat, which are different methods of analyzing multi-dimensional data.

A

I'll, probably still do that, but I'm gonna have to I've, never really been able to get my head wrapped around that topic, but I think it would be useful for people who are reading a lot of papers on like big big data papers and development and neuroscience where they talk about these methods.

A

But uh I'd you know, that's something like if, if you come up with an id and you put it on the board and then it's like, I don't want to do this right now. Well, keep pushing it off so, but I think that's so those are our major tasks for 2020..

A

I just you know, wanted to go over the board again to get people sort of recalling some of the things that we put on the board in the past and I think I'll I'll post those slides again on the project updates and on community building I'll put those in the devo worm channel on slack, and you can look those over and we can talk about it further if we have any other questions so now.

A

The final thing I want to talk about this, oh, should I ask, does anyone else have any comments or things to talk about before we move.

A

A

Okay, so the final thing I was gonna present on a couple of papers: we've done this in a while, so we've got some papers kind of building up here, so one of the things is going to talk about was sort of a coda to the the uh agent-based modeling paper that we've reviewed maybe two weeks ago, and um so agent-based modeling is a specific take on morphogenesis.

A

That is, you know you have these theories of morphogenesis and you can model them using agent-based models or you can understand them in the embryo and then come up with your own theories. And so that's these two papers, I'm going to talk about, are kind of going to discuss. uh This is a more you know biologically.

A

You know they don't really worry too much about like uh a mod. You know an insilico model or a computational model, they're just interested in the biology. You know, observing the biology and- and you know not even thinking about like the representation or anything like.

B

A

So this is james briscoe. This is a paper called understanding pattern, formation and embryos, experiment, theory and simulation, so the abstract is 25 years ago, lewis wolpert, the eminent developmental biologist, asked the question: do we understand development?

A

He concluded that such rapid progress had been made in the preceding two decades, and this was of course 25 years ago that it is not unreasonable to think that it will be eventually be known, that to program a computer and simulate some aspects of development.

A

This prediction has been fulfilled at least partially, with data-driven simulations of several different developmental processes being developed in the intervening years. Nevertheless, the question remains of whether we understand development and if simulations are sufficient to provide an explanation of development, so they're really interested in like taking the processes they observe it in nature and building a simulation of it and saying. Is that something that we really understand?

A

So you know whatever you see on your computer screen it that's neither here nor there it's like do we understand what that process is, because you know you can build like something like a pattern. Generator um in you know a basic pattern, generator that looks really like an embryo, but then you know it doesn't really behave like the embryo. The inputs are not at all the same, but they look very similar.

A

So that's kind of the idea um behind that statement now, while in silicon replications and models, are undoubtedly an important tool in the investigation and dissection developmental processes which complement traditional experimental methods. These need to be supplanted by theory that identifies principles and provides coherent explanations.

A

Here I use the example of pattern. Formation is excuse me.

C

A

In a pattern formation in the vertebrate neural tube to illustrate this idea, so they.

D

A

They say well, okay, now we need theory to understand the difference between like computer simulation and real biology, and that's kind of the perspective we've taken in this group is that we use theory quite heavily to understand some of these.

A

You know if we can create a a pattern or we can create pattern generation or we can analyze data, but we also need theory in the mix, and this is, I think, one of the tasks in our task board here was to create a theory layer for divalern, and so I think, that's uh that's why I put that task on there, because it's very useful to have a theoretical interpretation of a lot of this stuff.

A

So uh you know mentions that pattern. Formation and developing tissues relies on allocating naive cells and specific functional cell types and define spatial arrangements in temporal order. This is achieved by initially uncommitted, progenitors acquiring their fate in response to molecular signals that regulate the transcriptional programs that control cell functions.

A

So that means that you have uh these cells, that are, you, know, undifferentiated that have to take a fate and, and sometimes they can take multiple fates over development, and this is controlled by a genome inside the cell that determines the sort of what the fate is, and this is determined in turn by the interactions between the cells or their position.

A

We don't really know, but we know that those cells have to divide and then differentiate into different fates to become a functional organism, and so that's what we're dealing with how these programs are activated at the right time in place to produce a correctly patterned tissue is a central question.

A

These programs are encoded in the genome, and so we can use things like genetic regulatory networks or grns, and we haven't talked about grn's too much in the group, but this is a tool that we can use to model some of the expression of these different fates and, given you know, a model of the different genes and their expression, and everything like that.

A

So those are that's a tool we can bring to bear and that's very different from a pattern formation uh perspective like an agent-based model, because an agent-based model is concerned with the cells in their interactions, maybe their internal state.

A

But we don't talk about the internal state too much we just kind of like let this simulation run and watch the global pattern emerge, and so the grn is concerned with the individual cells what's going on within them, and so there is actually a workshop at um artificial life 2020 this past summer on uh genetic regulatory networks or modeling those networks.

A

um I can put the that video in the uh slack channel as well, so you can get a sense of what that looks like, but it's basically, where you have a bunch of genes that are you specify the genes, maybe with a binary string or you know some other mechanism, and then you express those genes at a certain degree of expression and then they're all linked together, so that one gene will one generalize and another gene to be active and you get these outputs that are then something that allows you to put together things in the phenotype so like two different cells might have different fates because their gene regulatory networks are uh different.

A

That they're regulating different gene the same genes differently, so you can use the same representation in all different all the cells in your simulation.

A

But if they're regulated differently, you get a different phenotype, and so it's it's a powerful approach um and you can do this modeling, and this is you know again. This is making your model more explicit or more biologically explicit, and you know that's something I think you're going to need to do. If you want to really understand development, but it requires a theory of what the genes are doing so a gene regulatory network is a theory of gene action.

A

It suggests that there are a bunch of genes that are working in a certain way and people of like you know people have done the empirical studies on this. So it's not just pulling it out of thin air, but you have this. Basically, this theory of gene action that you're using to build these models that determine sulfate and then, ultimately, the actual, what the organism looks like, and so they have some good citations in this paper.

A

uh Two ideas appear to be central uh first year and follow a modular design structure with sub-circuits within the grn at least partially segregated, so that their operation is relatively independent of the rest of the network.

A

So the grns form these modules that are, that have different functions that actually function within their own domain and they do their own thing.

A

So, within these sub-circuits network, motifs such as feedback and feed-forward loops, tend to be over-represented, and so these modules do their.

B

Own thing, but.

A

They have regular, you know they have like if you're interested in cybernetics, they have these regulatory actions, so two genes might sit next to one another, and one gene might affect the state of the other gene, but then there's a feedback that regulates that interaction, so that when there's you know, you produce a certain amount of output.

A

But then, when you hit a certain threshold of output, it brings things back down so that you don't produce too much of something, and you see that in nature, actually quite a bit they're different gene regulatory circuits that control like the production of enzymes or some sort of hormone that do this all the time they operate on feedback and when the amount of output reaches a threshold, it shuts the system down and then, when the output is depleted, it starts to ramp up again- and this is something that you know it's hard to model uh with just kind of like you know, um just sort of a simple feed forward model where you assume that all you know genes kind of act uh independently to produce something.

A

It's it's a very different view of like the thing that people were. Maybe if you think about in terms of genetic determinism, where, like each gene controls some part of the phenotype and then that's it, you know, that's not the way it works. You have these networks that are highly um non-linear, that produce all these regulatory outputs.

A

So it's you know this is, uh but people have done a lot of modeling on this now the question is is like: can we understand how this actually assembles a phenotype?

A

And the answer is sometimes in the case of spatially, restricted gene expression like producing this neural tube along the dorsal ventral axis you're, going to see this sort of thing where you can actually look at genes that have a certain that are expressed in a certain location along this axis, and you can actually model this very specifically and actually end up with a nice um anatomically specific um system of gene expression.

A

So there this paper goes quite a bit into grns um and then they also talk about dynamical systems and complexity. Theory um talking about recursive links, which is an idea of both positive and negative feedback results in the output of a system being recycled back into the system.

A

This leads to behavior that is not intuitive and is difficult to understand and predict, so we're using concepts from from dynamical systems theory as well um so yeah. So returning to lewis wilker's question do we understand development? Clearly, substantial progress has been made towards a molecular genetic and cellular understanding of development for specific tissues.

A

So we can look at the neural tube as a good example, but also we now have a framework in the set of tools that enable us to more candidly answer, wolford's question and explain the basis of her understanding, and so this is a nice paper if you're interested in geniux gene regulatory networks and maybe some of the dynamical systems tools that people have used to look at development, and so this is a nice paper. For that, then there's another paper that came up- uh and this is a little bit different.

A

uh This is from genes to shape during metamorphosis history.

A

So metamorphosis is this process where you go from, like an organism like a caterpillar to an organism like a butterfly now they're, they look very different, but it's the same species and what happens in development is that the there's a developmental process that produces the caterpillar and then there's this cocooning phase, where the caterpillar goes into the cocoon when it's when it's in the cocoon there's some decellularization.

A

That goes on some re uh recommitment of cells to different lineages, and then you end up with this butterfly that comes out of the cocoon, and so there's this process of growth and then sort of you know things kind of get reconstituted into this new form, and so there are actually two phases of growth in that developmental trajectory, and so in this paper they talk about.

A

uh How do you go from like gene expression to the expression of shape, and so this is interesting because it's historical actually and they talk about darcy thompson and conrad waddington. So these are two people who discussed a lot about um building. You know: how do you analyze a phenotype? How do you analyze growth and form?

A

um Waddington was interested in wing shape and drosophila, so in the it's, a fruit fly drosophila and they have these wings that uh have a diversity of shapes and it's controlled by it's famously known, to be controlled by developmental processes.

A

So when you have a mutant, uh phenotype or a mutant genotype, you can end up with a mutant, phenotype and they've actually been able to demonstrate how this works very clearly in drosophila with darcy thompson. I think we've talked about his work, where there is actually very interested in how mathematical laws work to um shape what we see in the phenotype.

A

So thinking about, like the self-organizing processes of soap, bubble, packings and hexagonal tilings, those are things that um actually explain some of the things you see in nature and in phenotypes, and so he did a lot of work on like looking at the mathematics of that, and um so we, you know we we can think about it in terms of morphogenesis, but we also have to think of in terms of evolution.

A

So if we just think about self-organization, we can think of these shapes, like honeycombs and tessellations and liquids, and these are things that you see not just in living things, but in like in inorganic things in nature, so you see it in water. You see it in other things that form packings and shapes, but you also, you know in living things. Of course, there's this evolutionary process, so this morphogenetic evolutionary process is also an operation. So these these shapes actually come from genes being expressed and genes that are changing their frequency throughout evolution.

A

And so when we look at genetic control and wing development, we have to consider not only the pattern but the genes that are underlying that.

A

So this is an example here of uh columnar cuboidal cell cha. She uh cell shape change in wing morphogenesis, so this is uh fristrum uh diane fristrom in 1969 did these drawings of the wing, and so this is wing morphogenesis how it's changing shape from this columnar orientation to this cuboidal orientation and you know a lot of the early um anatomists. You know they had to draw things out by hand and that's actually very useful.

A

Even today for understanding how these things change in development, so looking at like, if you draw things very carefully, you can see a lot of the shape changes, and so this is actually something you'll find a lot of in developmental biology, and so this is a good source of data. It's hard to maybe get a handle on how to quantify this. But um that's that's where the data comes from. You know. That's that's kind of the early passes at the data, so we talk about cell tracking here and you can get.

A

You could get something like this from cell tracking, but a lot of these images going to give the sense of what's actually happening in terms of the process.

A

And so then, this is actually from a different paper, and this is this is a sort of a model where they've taken these drawings and they've created a model of like a compartmental model here of these type of processes like cell elongation and so for cell rearrangement, where the cells are just rearranged in development and changes in cell shape, where the cells actually change shape, and so there's a lot of good stuff. In this paper, I'm running.

E

Out of time here to.

A

Get in any more of it, but if you want to look at this paper as well as the other paper, it's in this google drive, folder and I'll put the link up later, but I'll put it in the chat right now, so you can go.

B

Into that folder.

A

And then I'll put the presentations up on in the slack open room slack as well as the links to the periodicity paper.

A

So any other comments before we go any update on the boring billion. I'm actually going to talk about that with someone later today. So we'll probably have some updates this week sometime.

C

um Obviously that looks interesting and I I'm sorting out some of my.

B

Plan of attack, but yeah, I'm probably interested in.

C

The theory component of uh the deep learning and other issues like that so more about that later.

A

Okay, yeah, we can work reorganize these sort of tasks. You know I mean I think that there are ways that we can contribute we're kind of like throwing out ideas, but if we can get uh you know better handle what they involve, then we can get people, you know and get things moving. So anything you. Any suggestions you have on that front would be good.

A

Okay, yeah thanks for attending this week, and I hope you have a good week and again we're meeting- or you know this at this time from now on this fall. So I know there was a time shift and that was on me. So thank you for being accommodating and um see you next week.

A