DevoWorm Lab Meetings, 25 Sep 2023

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: DevoWorm (2023, Meeting #34): Hacktoberfest plans, Enumerating human cell types, Morphogenetic flows

Description

Planning for Hacktoberfest project (annotating embryo data). What are the number of cell types in a human body? Cell size/function scaling and type distributions via meta-analysis. Dynamic morphogenesis and morphogenetic flows: reaction-diffusion, discrete computation, and circulation as a means to break and reinforce symmetry. Attendees: Sushmanth Reddy Mereddy, Bradly Alicea, Jesse Parent, Susan Crawford-Young, and Richard Gordon.

A

Hello, how are you.

B

A

C

A

Just with your graduate uh program, yeah.

B

Just with the graduate program, they're being fussy or something or you know, didn't like what I did, how do they know what I did I didn't? um I didn't haven't handed anything in, so I decided to write it up. Oh yes, so I have 5 000 words for them right now. Okay and another couple thousand words, maybe I can hand them a ten thousand word essay and see say this is what I've done is ten thousand words. If you want more paper.

A

B

Yeah, it's five thousand words and it's definitely not finished it. Oh.

A

Yeah I sent you an email with those papers. I don't know if those are the right ones.

B

Well, one less than the other one looks interesting. So, okay, it's it's good! That I yeah I was just trying to get at how things tiled together and.

A

Good luck with that I don't know. Usually you know.

B

It's it's just annoying like you're gonna, put more pressure on me like I I, can't.

A

Thank you, yeah I,.

B

Can't work any faster than what I'm doing. Thank you, yeah.

A

It's not you, can you just uh pop it out like in a couple days.

B

No, and and the other thing is, is I'm good at writing. So if you want output best way for me to do, that is to write and write to them. 10 000 words.

C

B

It is so you go ahead, yeah sort of Revenge of the Nerds. Let's write this up.

A

That's good well Jesse, little dick. How are you guys doing.

B

Okay, the students, the students from India.

B

A

Sushma was going to join he's going to talk about uh October, something called Oktoberfest, which we've.

C

A

In the past, but it's uh something that involves the GitHub repositories and things like that: I, don't know if he's gonna be here or if he's not going to be here, he sent me a okay, Jesse, so hopefully be listening. It did I just because sushma sent me a message just recently. uh Yeah so well, I'll talk about it and uh I. Don't know.

A

Maybe he'll come in um so yeah, so one of the things we'd like to do in the next. Well, you know imminently I, guess, because it's almost October is to host something called hacktoberfest. So Oktoberfest is a time that you sort of an official period for GitHub, where they like people, expect to be able to contribute to GitHub repositories and some organizations offer a few have five commits during that month or five pull requests that you get it some sort of uh recognition or free, swag or which is like a t-shirt or something like that.

A

Oh there he is so I, don't know if we don't have the bandwidth uh for swag I, don't think okay camera, but you know that it's it'll be a nice thing to participate in, uh especially we've done. This before we did this I think two years ago, or maybe even last year, I can't remember, and you know we had a pretty good response, so people will come by the repository in this case. It's the GitHub organization.

A

Look it over, they might. We have to create some issues for it first and then uh people will. We can advertise those issues and then, um in the meeting, for example- and uh we can have people work on them and they can get some sort of recognition for it. So sushma, hello.

A

D

Understand yeah.

A

Yeah, that's what we're talking about. I was just setting it up for you.

D

um Okay, uh um this is the project. I was thinking.

C

D

Has a project: actually, it is off fine-tuning the data front door modem, but before fine tuning we need the data set. I was thinking to create a some data set for us. I already may put the tip file from the cell tracking Channel and data set and converted them into images.

D

Right now our plan would be creating a custom data set I mean everyone need to learn the cell image analysis according to position. They need to name the each using a tool called vmix, and since you can, if you can play this video and can I share my screen.

A

A

um Not yet not sure why it's not letting you share.

D

Yesterday, okay yeah.

A

Where would we let him uh rejoin see if he can log in and uh share his screen.

C

C

A

C

A

He just said, don't know well.

C

D

This video, actually you.

D

This is a custom data set, he was making using a open source. Library called we make sense dot AI. It is like free and open source.

C

Oh can't you see the string.

A

D

This is the website I'm talking about you can upload all our images right now what we have and we make them into a c object. Detection uh kind of, if you see in this video.

A

Yeah, you can see it.

D

These guys are making like a ground truth of it, so we are working with ACL games right. uh We will use this tool and we will annotate its name for lineage analysis like it will be custom object, detection plus lineage analysis model at the end.

D

If we fine-tune this model on our on the data set, whatever we are making for hacktoberfest for next gsop, it could be a combined model, uh it will segment the cell also and it will we can get the positions of each cell and we can also get the what we call the lineage analysis in biased model.

D

Actually, we have three different models for each thing. So if, by using this, we can directly segment the cells and but in Oktoberfest point of view, we'll just make the data set. Because right now we have around 5000 images of uh cell CL games. We need a custom data, so I was thinking to give the whoever participating hacked over first give them this data set and try to undertake them. That would be the October 1st activity after annotating. We couldn't extend this project if we can complete this data set on time.

D

In hack cover first, they can start working on the model also or it will be next use of project. Finally, it's an object, detection segmentation and naming also Bradley before uh you who taking any decision. Please give a look on this video. This can be not only used for uh C Elegance, but also in other kind of organisms.

A

Yeah, so this is, this could be some kind of no finish up. I was gonna, say something, but.

D

Yeah, it could be perfect, it's a large, well-trained model, so it could segment The Sims properly. He just uh tried to fine tune this model for this video for only for 15 or 25 images, and it is giving best results at the end. So if we make our custom data set of 5000 images, fine tune, it could be act as a best model for cl games. I think so. Yeah.

A

Yeah, that would be great, um so you you have, you said you have a data set in mind or.

D

Yeah, yes, I have a data; actually, uh maybe I I will provide you. The data set images. I have converted all the active files into images.

C

D

They are graceful images. This is how they look. Pretty I will show you.

C

D

So if you see Bradley, you can see them.

A

D

You can name them using the phone, so we can just use for object detection. Actually, he also have the microscopy images also when he was annotating. It.

D

Just name them at least create a boundary for it by using this tool and name them. Each option.

D

Sometimes are overlapping over this thing, so go for this before yeah, they are not colliding actually see yeah.

A

Somewhere harder to so this is about like basically having to annotate these or with with some data.

D

Annotating, it.

D

Before, please give a look on this video I hope that would be helpful, for you also share the video link over here. Yeah.

A

D

We can decide, we can continue.

A

And so if people are going to do this, what do we need to provide them to like yeah.

D

Nothing with the whatever software we are using is open. Source makes sense, dot AI. Now we need to divide the work. Actually, we are just participating and give them the data set and whoever works and companies we can issue them swags. Maybe we can give a 50 images to each person economically.

A

Yeah yeah and yeah, so it would be like and then there'd be like some sort of.

D

D

Collaborate in the development projects- they could just work with this at the starting point, so we can evaluate them who are interested in the project who are not interested in them. Maybe if we can reach out by next piece of time it could be you. It could be helpful for you to filter out the movies like genuine and also.

A

Yeah and of course, you need to have a way to submit it to the to a centralized repository, I guess: I, don't know if you could just put it on to use GitHub and then you'd also need some weight away cross check, because people.

D

May not I will create the readme file for this. Okay and I will post it on the drive for the data set kind of thing and then provide the link uh download the data they can just download. So.

C

A

Yeah and then, of course, there will be like a cross check, because you know unless well I guess you could just provide people with a template for like submitting information like, for example, you know have like a format for the position or the name actually.

D

It is completely mentioned in the video of Bradley. They can see the video in this tutorial and they can annotate it.

A

All right, yeah, I, guess I- don't understand that right now, but.

D

Yeah yeah: that's this all should be like a thesis kind of thing. Once you watch this video, you will get a complete idea. uh Whatever we can discuss is tomorrow after your watch.

A

Okay, yeah, that sounds good.

D

I, don't slack acting so.

D

A

Yeah, that's that's fine um yeah, so this looks good. This looks good.

D

Before this watch this video, if you have any other model organisms in your mind, to fine tune it as two as a project. That would be great also because this works with any modern organism not only CL yes, but sale against, because I know about it. Other model organisms- I, don't know we can give the synonyms for each and, according to the position, wow.

C

D

That's what I was thinking if you have anything in another thing in mind, you can work with this kind of stuff. We can definitely watch it.

A

Yeah we actually, we do have Axolotl data. That season is provided another. You know open I mean they're all with different C elegans. There is enough, there isn't a nomenclature that you can attach like C elegans. There's like uh you know, every cell has a nomenclature, identity and you know okay and then, but with the Axolotl embryos. I guess they're a little bit different, but they can still. You know you can annotate the cells sell by sell. So it's not like it's it's impossible or it might have. There might be some value in doing that.

A

So I'll look I'll, see I'll, think about it in the context of how the um how the software works.

C

Okay, whatever.

A

It's great thank you. I'm.

B

Just wondering if um thick as you um upload the images from the reason to ask a little batch that I imaged with the ball microscope, um I I tried to do that and it didn't work and I still don't know. Why.

B

I I gave them a disc Tim, physically USB yeah. This can with the images. So he does have them it's just, but I and I did try to upload them and it didn't work. So I don't know. What's going on in the sink.

A

C

B

Yeah I do have some some images that that would work um there um distorted as an oval on in some instances and they have Shadows from the microscopes on some of the images.

B

um So you're, working with with things like that right, at least I have I, have some but and um I thought of maybe putting some some more light around the microscope to get rid of the Shadows.

B

But I need to find some more salamanders willing to lay eggs.

A

Yeah yeah, okay,.

C

So Susan I've had two problems: lately uh uh the upgraded sink and the new version didn't work on my computer.

C

Oh fine, and it took a couple of weeks to fix that uh if I'm back to the old version- and the second is that I've exceeded the two terabytes of my internal disk- oh okay, okay, so I still have your sticks, but that's it.

B

Okay, all right, there's um anyway. Okay, there are images.

C

B

Just a lot of data which makes them hard to handle.

A

Yeah yeah well, yeah I mean we can I have some data that you sent me and I have that in a separate repository. Maybe we could just use that that would be like a good first pass because I don't know, um we didn't have to work out. A methodology. I also been talking to Hare Krishna, and he was mentioning that he wanted to have like a I.

A

Don't know he was talking about a reference frame for the images and I said that you know that I don't know what he was referring to, uh but it hadn't been worked out, so he was gonna. Try an automated method of that and I'm not sure what he's referring to, but I guess like I'm, not really sure what he means by that. But um I just didn't think that yeah. That was something we had. So you might be able to provide something like that. So.

B

Rather, should I send you a stick with the images on it.

A

uh Well, I have yeah, you could do that. I have some. You sent me a.

B

Stick a while, oh.

A

Okay, yeah I, don't have.

B

Those either so that I'm sort of annoyed at myself for not getting there, but my computer upstairs works now because they installed fiber optic and proper system. So I can now now get onto the internet. There right.

C

B

That, oh, that helps that helps the computer operate correctly um anyway, so yeah, my upstairs computers now Okay, so more accents to images. Yeah.

A

B

A

B

I need a new one.

A

We can work on the Axolotl, maybe as well so that yeah, that would be good, uh we'll set that up. I'll, take a look and we'll talk about that offline, so yeah and then the idea would be we'd have.

A

Let me see if I can pull up hacktoberfest from a I guess, maybe two years ago, I'm not really sure.

A

So this is the Diva learn uh organization. uh We have people, learn the repository and we were doing hacktoberfest in this Repository.

A

Well, okay, we have so what you basically do for uh pactober festas you have, the Repository usually put a badge on it. I think we have the badge here. I, don't think we have it right now, but we had a badge here. You can make a badge and put it on, um and you know the idea would be you'd advertise this in places where people are likely to.

A

You know be interested, so it'd be like uh on social media or in you know you might talk to some people in a university or something and they as long as they know where it is, they can come here. They can look at the repository. Look at the readme there's a badge usually saying that you're participating in hacktoberfest, because people sometimes will come from you know.

A

Sometimes they'll peruse get help to look for open, acto, professed repositories, and so once that happens, they you know can go in and look it over and then one way to sort of facilitate a lot of these things is to set up a set of issues specific to hacktoberfest.

A

So I'm not sure if we have any better from previous years. Okay, here we go so when we did this. uh This was actually in 2020 uh these examples, but we had uh a couple of issues here and we have this label hacktoberfest.

A

So this is where you put this label on the issues, and you know it tells people. This is what we want to achieve during Oktoberfest.

A

So we create this set of issues and we give it a description and then you know we give them instructions and what they can do. So here we had my yoke set up an issue here and uh then Ravi said: I would like to do this and then my ex is we're looking forward to a pull request.

A

Another person who's interested in contributing to this issue Ravi then issued a pull request that got merged successfully and then I guess this is uh different person yeah. This is a different person. Thank you for your interest. This is the second person. I would like you to look at contributing MD, which is a file on the repository and checked in arms used for a proper commit message. So you know you have you have to coordinate people's requests and all that, but basically that's how it works.

A

And then, if someone commits like I think the general convention is, if they make five commits during the month, they can get like I think they can get a free t-shirt from uh from GitHub itself. But usually we offer things and, like I said I, don't know what kind of position we're going to offer things. uh But you know I know that, like recognition is one thing, but um you know that that's that's less important right now. The important thing is we get the uh details together.

A

We can just tell people what to do and how to do it, and then we can even create like a new uh directory for that. So in the Repository, so the devil learned in the organization we have the diva learn Repository.

A

We have other repositories like data science, demos devograph. We can just create a new repository here. That's you know. Basically uh I, don't I don't want to call it over first I just want to call it like. Maybe like uh whatever you know, I don't know we'll have to find a suitable name, but basically yeah.

D

Create a repository called data sets and whatever I have the data set. I will upload over here. If you can download it and annotate it, uh we will create like uh issues in the GitHub page 1 to 50 images.

D

uh One issue 5200 I mean for some people like this people can just go and visit it and Ascend themselves to that issue and work on that right now. I have another thing in my mind and uh sure in my mind, I will find it over the owner. I will take permission from my is that okay, yeah.

A

That's good yeah, that's good yeah, so we'll create some issues, sort of to start with, and then people can also sign. Well, they're gonna assign themselves the issues. They, of course, can create their own issues but yeah the way it usually works. Is you start with a set of issues.

D

Yeah, okay, I'll create the data, sets repository or development and I will try to uh create a structure one and after the approval from Mayu. Can you uh we'll make it as public yeah.

C

D

People can just come and just take up the issues and send to them and work on it.

A

Yeah that would be great, um I have a yeah I have so and we have materials. I have to go back and look at it, but we have some materials for promoting hacktoberfest.

A

So we have like a badge that we use for the orc and we have you know other things that we can and then you know we want to get it get the word out um and so that that's good. We also want to use the hacktoberfest uh label for the issues, because we want people to be able to find them on GitHub and I. Don't know if well, the hacktoberfest label is in the Bible worm repository. You may have to recreate it for this new one, but that's that's trivial.

C

All right, yeah.

D

Okay, we will discuss it further in this coming week and we'll try to complete it by October 1st, so we can work. Everyone can work on this.

A

Yeah yeah yeah, so in past years we've had quite a few people contribute we've had like I think one year we had like 11 or 12 people, so that's good, a nice nice pipeline into that work and then, of course, this is something that people will use for like.

A

uh We can actually advertise in fact and say that this will be part of next year's gsoc. So.

D

People have extra incentives, yeah yeah people will show extra interesting.

C

D

Yeah, let's see what we can.

A

Do yeah that'd be great, so yeah, that's great, uh so we're getting somewhat close to the end of uh gsoc official. Well, the extended period so uh both sushmath and amanshu have been extended to the 24th of October, so coming up uh within about a month is the final submission for gsoc and a few words on that. um So you know when you're working in a project oftentimes you have to pick a point at which you publish where you submit uh some version of something, and so you know.

D

Yeah we are working on it. Actually, I was working on the paper. Most of the coding part has been run, I need to restructure it and make a documentation for it. That's what was left now right now. uh I will work this couple of weeks, maybe it'll be completed, I'm just not concentrated in last couple of weeks, yeah. So next four weeks we have right exactly or it will be complete. Okay,.

A

Yeah so I guess I guess my point there was that at some point you know we always make like the the version that we release or we submit, and so uh we have I guess the way that uh Google wants to do. This is to have like a submission like a readme or something that you can, that they can read through and they can look at the submission and be able to you know actually download or run the software in some way.

A

Now the program has expanded a little bit this year to include things that are not software, just basic way to have a place. That's going to be a permanent home for the submission that has like a readme, that's very clear, to read and so forth. So um that's good I! You know, I can provide more details on that. If you need, but just to let such make them.

D

Aware of that, yeah.

A

C

A

Okay, so uh why don't we switch gears here and talk a little bit about uh some things, I actually uh like to talk about some things on cell types and estimating cell type number and some new work. That's come out on that so now, I'm going to share my screen and.

A

So there have been a number actually I, think two papers that have come out or actually one paper- that's come out and then some other things in support of that that have kind of thought rethought. This idea of counting cells and cell types in an organism, and so we've talked about this in past meetings.

A

uh John Bonner I think published a book about 35 years ago, where he was trying to make estimates of cell number and cell type in the human body and more recently, people have made estimates of cell number and cell type, but they've been wildly variable, and one of the reasons is because, what's the Criterion that's used, you know what is the Criterion? People are using to determine what the cell type is and that's no easy thing to to.

A

You know sort of make a Criterion for us, because you know there are all sorts of ways you can classify a cell, you can use phenotype, you could use function and more recently, people started to use molecular markers, but molecular markers are pretty Broad in terms of like you know what to include what to exclude. Sometimes you have markers that most cells have sometimes you have markers that are restricted to certain cell types, uh but you know it's it's hard to really kind of support that with other types of features.

A

In other words, you might have a marker, that's very specific to a certain sort of precursor cell, with the morphologies of those cells might be Divergent. So that's very hard to do, and so uh there's this paper that came out recently in pnas, and this paper is called the human cell count size distribution, so the authors are listed here, looks like they're people from Stanford and and other groups, um and so they've done this in uh 2023 version of this, so I'll read from the significant statement that I'll read from the abstract.

A

So the significant statement here is a consistent and comprehensive quantitative framework of the cells in the human body, and so that's what we're trying to do we're trying to build a quantitative framework, not just cell counts, we're cell type counts, although we can do that as well.

A

That's useful, for the reasons why we just talked about um doing machine learning and deep learning on microscopy images and to do that you need to have Netherlands segmentation uh worked out, but also sort of differentiating between types of cells in the images, so that that's where it's useful, but having a larger quantitative framework, it could benefit many areas of biology.

A

We compiled data to estimate cell Mass size range and cell count from some 1200 cell groups from the smallest red blood cells to the largest muscle fibers. So these are all sorts of different cell types. Cell morphologies gene expression profiles Etc.

A

This ranges across 60 tissues in a representative, male represented a female and ten-year-old child in the gender isn't specified there. We find large-scale patterns revealing that both cellular biomass in any given logarithmic cell size class and the coefficient of cell size variation are both approximately independent of cell size.

A

So we're talking about like a logarithmic, transform and a coefficient of cell size variation. Those are the two measures they're using and those are independent of cell size, so you can have actually information about the class and a measure, variation that are independent of the size itself. So, as the cell gets bigger, these things are independent.

A

These patterns are suggestive of a whole organism, trade-off between cell size and count, which means that you have the number of cells the size of the cells. Those things are: there's a trade-off between those and imply the existence of cell size homeostasis across cell types.

A

So this is where we talk about. You know how the cell is regulating itself, maybe its identity, maybe its size just overall and how that's regulated across different cell types. So if you have a certain role as a cell, uh you have a certain identity that size of limit. You know, it'll limit. What size you can be with the optimal size is for that function Etc.

A

So this is the abstract um cell, size and cell count are adaptively, regulated and intimately linked to growth and function. Yet, despite their widespread relevance, the relation between cell size and count has never been formally examined over the whole human body. Now, what they mean by that is that there hasn't been like a high throughput data set generated.

A

People have made estimates, but there hasn't been this I guess what they're doing is they're doing like a high throughput of analysis of this. So uh here we compile comprehensive data set of cell size and count overall major cell types. The data drawn from over 1500 published sources. So this is sort of almost like a meta-analysis.

A

And so this this involves a huge number of sources of data. We consider the body of a representative male 70 kilograms, which allows for further estimates of a female 60 kilograms and a 10 year old child 32 kilograms.

A

We build a hierarchical interface for the cellular organization of the body and they have this app here. The human cell tree map giving easy access to data methods and sources.

A

In total, we estimate total body counts of the approximately 36 trillion cells in the mail, approximately 28 trillion cells of the female and approximately 17 trillion cells in the child.

A

And these you know there could be different cell types in that count, but that's their estimate. These data reveal a surprising inverse relation between cell size and count applying a trade-off between these variables.

A

So if you, the basically the larger your cell size, the smaller the number of cells and vice versa, which makes sense- if you think about like uh you know, if you think of the body as like this trophic system, where you have cell types with these functions and trophic functions, and they have this sort of relationship between their size and their function and their proliferation ability or the need to proliferate.

A

So if you're, a blood cell red blood cell whey, blood cell, there are many more of you than say, if you're, like uh cell on the body that you know like a neuron or something, um and so these data reveal a surprising inverse relation implying a trade-off between these variables, such that all cells within a given logarithmic size class. So they they transform. This into logarithmic size classes, contribute to an equal fraction of the body's total cellular biomass.

A

So that means that you have this, these fractions of biomass. In them, the size classes are roughly equivalent, so the smaller cells are more of them, but they contribute in proportional numbers.

A

We also find that the coefficient of variation is approximately independent of mean cell size, implying the existence of cell size regulation across all types. So there's a cell size regulation you get to a certain size, you stop growing and there's this the size uh limit. You know, and so again this may be energetic. This may be something to do with function.

A

This may just be something to do this, with kind of what's uh uh sort of the number of cells that are already in the body, these sorts of things, our data served to establish a holistic, quantitative framework for the cells of the human body and highlight large-scale patterns in cell biology.

A

Let me see there was a comment in the chat. It says the size bumper trade-off is similar to that a public plate axolotls the other size number trade-off yeah, so yeah probably played axolotls, have the same sort of relationship.

B

How many um cell types are there? Did they actually come out and say that.

A

uh Let's see they say: 1200 cell groups, I, guess that's what their types are so yeah they don't. Let's see if they give a number I mean. These are I, think largely estimates I mean they do the counts, but, like you know, you can't observe all cells necessarily you're, just getting the cells that they can have. They have access to.

A

So here they talk about cell types, uh one level. It is apparent that the body's cells are the right size to best perform their function. Irregularities in cell size are often a pathological sign of disease or a marker malignancy. This means that you know cells can be good on a cancer pathway and start to you know, uh show all sorts of aberrant, phenotypes, and so that they're, you know technically they're that origin is from like one cell type or another cell type.

A

So this is where you know we kind of want to filter those out, but we also have variation, especially in cells like fibroblasts, or something like that, where cells self uh shape variations where size variation can happen.

A

uh Moreover, changes in cell sizes are associated with changes in biosynthetic capacity and metabolic function, especially when cell size chains are not accompanied by changes employee, which is the number of chromosomes in the cell cell types, such as myocytes, neurons and adipocytes, which are fat, cells, myocytes or muscle cells very extensively in size, but each cell has a size, specificity suited to its function.

A

So you know you have these types that vary in size, sometimes within the category, but each cell has a size, specificity, that's suited to its function, so, for example, muscle cells and sensory cells are smaller in the face than the legs, so you have muscles in the face that are very small. There are water muscles in the face in the human face and in the legs you have larger muscles that span like the entire limb or the entire segment one segment of the limb.

A

So you know these are just there's a relationship between sort of nerves, that innervate like say the face versus the parts of the leg and maybe the muscle size, and so we know that from developmental neurobiology. So you know there are these relationships with the nervous system, relationships with the morphology and the need to have muscles, or you know, sensory Fidelity in certain parts of the body over others.

A

So this size homeostasis is thought to be 19 to some degree across all cell types, but the strength of size regulation, which is often measured as the coefficient of variation in size, remains unknown across the full cell size range.

A

So uh at the high at a higher level of organization, the overall functional capacity of a tissue depends on its total biomass. So this is again, you know the tissue. If it's biomass, if it's larger, you know its function is different than if it's smaller um since or what it can do. Since the cell types of Any Given type are relatively well constrained, the tissue's functional capacity is determined principally by the total number of its cells, so the total number of its cells determines its size. The size determines functional capacity, and so that's that's.

A

How they're sort of characterizing that uh two important studies in the past decade has surveyed the total number of cells over the largest tissues in the human body? The total counts converging on values of 30 to 37 trillion human cells. That's 16 and 17, which I'll get to in a minute, with an approximately equal number of bacterial cells, so cell size and number of direct Coral areas of growth and proliferation and among the most fundamental quantitative attributes of lace, basic units, the relationship between cell size and number have never formally been examined body.

A

So they do this sort of they compile these size abundance of diversity. Statistics, um let's see 16 and 17..

A

These other studies that they're, citing so we'll get to those figures in a minute that I just in a glossed past, but 16 is uh be on Coney at all an estimation of the number of cells in the human body, and this is from 2013..

A

So this is from annual uh annals of human biology in 2013..

A

So that is one paper. That's attempted this in the you know, maybe within the last 10 years put that in the chat.

A

And then we have uh 17, which is Senator Fuchs and Milo, revised estimates for the number of human and bacterial cells in the body. This is where we're looking at uh human cell, bacterial symbiosis, but also the number of bacterial cells and uh living in in relation to the human cells. So this is plus biology in 2016..

A

So there are a lot of really interesting citations in this paper. Actually now, if I look at it, um so I'll post this or I'll send this paper out by email and post it in the slack, and you can look at it more at the reference list more, which might be very enlightening in and of itself.

A

But uh let's go up to this, so this isn't, let's see so this is an image of different cell types, so we have a platelet cells which are very small endothelial cells, stem cells, which are you know, a little bigger, fibroblasts, of course, which have this morphology that branches out, but isn't like a neuron with long processes. It just kind of has these looks like kind of a star or something like that, and you know they can have variable shapes.

A

Erythrocytes neutrophils alveolar they're in the lung, alveolar hepatocytes neurons. Of course this is a one example of a neuron. There are many different examples of neurons again uh very complex processes, adipocytes which are fat cells and then myocytes, which are muscle cells which are centricial cells.

A

So there are all sorts of cells that we have in the body, different sizes, different tissues and they play different roles in the body, and so you know this is where we get this above. We get these different cell types, color coded onto these bins that they talked about. So these are the bins where you're taking cell mass as powers of negative powers of 10. So you get like negative grams, so 10 to the negative fourth grams 10 to the negative 6 grams 10 to the negative 8 grams 10 to the negative 10 grams. So.

D

A

Very small and they're we don't weigh very much in and of themselves, but in the aggregate they're putting together these bins to show that the cell count goes up as the cells get smaller. So you can look at the size in terms of mass instead of looking at in terms of just measuring the size.

A

By looking at an outline and looking at the inside we're just looking at mass and then we're looking at cell count, the cell count goes up with mat reductions in mass, and this is roughly I guess logarithmic, and this is the equation here.

A

So this is actually another really interesting table. This is where cell class distributions are sort of co-located with Selective tissues. So we have all these different types of cells, lymphoid, non-nucleated, blood, granulocyte, other blood cells. We have myocytes cardiomyocytes, striated myocytes, we have endothelial cells, epithelial cells, stem cells, parasites, germ cells, neurons, fibroblasts, glial cells, osteoids and adipocytes.

A

They all have these different color codes and so up here what they did is they put this into different tissue types or at least they collect them for different tissue types. So you have the cell count, which is an estimate, but it's basically a really large number in each tissue, and then you have cell biomass and grams, and then you have these different sort of fractions of each for each tissue.

A

So for the testes, for example, you have this light green, which dominates the count, but maybe is a little less prevalent in the biomass, the light, green being germ cells, of course, and then you have a darker green band, which is which are stem cells, and then you have other. So you can see that in each tissue you have these different cell types. They have different roles to play.

A

um You also have the central nervous system, peripheral nervous system, adrenal, glands, liver, intestines and so forth. So you can see that the cell type variation across those and the intestines. You have a lot of blood cells in the skin. You don't have as many blood cells. You have a lot of epithelial cells and endothelial cells.

A

uh You don't have as many endothelial cells, but in the eyes, for example, you have more neurons, so we can go across the different cell type or the different tissues and look at the distributions. That way, uh then there's this other figure here, which is figure one.

A

This again has these different cell types, so they give little pictures of them and they give like a color coding, and so this is actually a visualization where they put things in a sphere and they like take a fraction of the sphere and they have like a percentage. So it's like a way to represent different categories within categories.

A

So, for example, you have all blood cells here in the red and within all blood cells, white blood cells, nucleated, red blood cells and then within those you have lymphocyte T cells and neutrophils lymphocyte, B cells, and over you know, so you have these categories of them categories.

A

You know, fibroblasts, you have stem cells, endothelial and epithelial cells. So you see that there's this for the cell count, you have these different types of cells, you have a preponderance of of white blood cells um and that's in terms of cell counts.

A

So that's different from biomass, which is in B, so cell counts as an a biomass is in B biomass, of course, by by weight instead of by count adipocytes dominate or fat cells, and so you can see that you know you have a lot of fatty tissue in your body in terms of biomass, but that's not represented by proliferation of cells. The number of cells that are adipocytes is rather small, but it contributes to a large amount of biomass.

A

Conversely, your white blood cells- your you know, the cell counts- are dominated by white blood cells are very small cells that predominate. They proliferate a lot but they're, not that they don't contribute that much to the biomass.

A

So it's a a trade-off there, where you have, you know, weight of the cells or what you know what those cells contribute to tissues into the body uh versus the number of cells. So you know there are a lot of reasons why you'd have a lot of cells. uh You know white blood cells function is you know? Sometimes they need to be replaced they're. They need to be all over the body.

A

There are a lot of reasons why you need a lot of white blood cells, but but at the same time you don't need they don't need to be very large to do their job. So when you have like you know a predominance of those kind of cells in the body relative to other types of cells, they actually don't weigh that much. So that's what this graph is all about.

A

Let me look at this link here in the abstract. So this is the human cell tree map. This is the tool that they have put together for this. So this is a app that you can open up on your desktop.

A

This is that figure and it kind of shows Us distribution. Naturally, you can explore it in this case, so we have the human cell tree map. So this is a tree map representation and basically it's like a tree that they visualize the different categories by size. So it breaks things down hierarchically into categories and subcategories, and then you can explore the different cell types here. So in granulocytes, for example, there are a number of different cell types.

A

This is both by counter by biomass, and so you have this graph here. The count on the right you have cell Mass, which is in grams. You have cell count, which is a positive uh powers of 10 in some assets, fractions of of one. So it's like you get one e negative, four winning negative 11, that's the range so they're in grams, so it's a pretty they're pretty light individually, but uh and then cell count is.

A

You know up to power uh 10 to the 13th, so we have a lot of we can put together and again. These are estimates, but we can put together different counts for different uh collections of cell types and we can get estimates that way, so they have that all plotted out for count and then for biomass. This is done: uh cell Mass versus Cell, biomass and grams, so you're, just looking at the mass of cells versus the biomass, which I assume is the tissue or the organism, I guess the tissue.

A

So you have platelets erythrocytes, you know, so you have all these different, uh it's all types of different weights and then the cell Mass. So uh this- and this goes through some of the different organ systems.

A

So this is basically the tool and this would be nice to play with again the uh link for that is in the paper abstract.

A

All right, so I have a couple of figures here. This is just a close-up of that figure one. So this shows this in A and B. Of course we went over this. uh The second figure is into this figure. Three I think that we looked at where they have this logarithmic uh uh regression between cell count and cell mass and the different cell types yeah. This is so this is the cell counts in humans.

A

uh I wanted to go over that paper and some of the figures, because I think it's especially what we've talked about in the meeting uh in previous meetings. We've talked about this topic before and you know it's kind of like we've gone through the literature and it's been very hand wavy in terms of the types of types and numbers of cells and there's a good reason for that. But you know this has been.

A

This is I, think is a it's it's from what I can tell a meta-analysis where they look at some of the features of the cells. I, don't think they actually use any uh next-gen sequencing in this, but you know at least they don't claim to use the markers as any sort of identity. So they just go through the literature and look at what people have kind of come up with.

A

But that's that's what we have state of the art in that area. I guess.

A

I guess any comments or questions about that.

B

Well, I think that my modeling of epithelial tissue only would be a good idea. There's too many cell types.

A

Well, yeah, that's I mean in modeling. That's an interesting question like what do we want to focus on in terms of cell type proliferation, because there are different Specialties? You know. Different organisms are different specialization, so this is just inhuman. So, like Nick mentioned that you know in Axolotl, you can draw parallels but like we don't have uh good I mean maybe in C elegans. We can do this kind of exercise where we look at like because the cell types of course they're.

A

um You know they're they're, you tellic organisms, so they have the same number of cell types in every individual. So you have this in the in the lineage tree, of course, is deterministic, so we know kind of where we're going to end up. We know what the tissues are, but we could actually build that distribution pretty quickly for C elegans, but for other organisms like Axolotl, you know it's going to be different. You're going to have different types of tissues.

A

The tissues are going to be composed of different cell types and that's largely based on functions so where you have functional uh requirements that differ, you'll have different distributions of cell types, but like in terms of modeling, you know we need to model every potential cell type, I, don't know, I mean I. Don't necessarily think we do. Although you know modeling is an exercise in like parameterization right, like you figure out the range of values you need. You don't know exactly what the cell state is. That's not like.

A

Maybe that's not the point of your exercise, the exercises to see how basically a cell like that, will behave in under certain conditions. So you know it like you know. I have to stress that, like cell types are very fluid in a lot of ways like.

C

A

A lot of variation around the edges of those categories- they're not necessarily nice new categories. Sometimes they are but oftentimes they're, not especially like across development or, like you know, across people. You know you could have cells in various. uh You know states of life like some. Sometimes they die. You know sometimes they're apatosing, sometimes they're going down the cancer route. Sometimes they're. You know interacting with other cells,.

B

A

Yeah yeah there was a lot of work on like wound healing that people have done like both in the biophysics and in the lake.

A

You know identifying which cells are involved in which processes, because that's pretty a pretty complex process where they have to migrate into the wound area and form a matrix that allows the womb to be closed and even, if you know, morphogenesis at that site. So it's a pretty interesting area.

B

Yeah yeah, it is yeah.

A

So if we have anything any comments or questions before we go.

A

C

A

I said we'll be working on hacktoberfest for those interested uh that'll probably be up and running by the first of the month at least. We hope so I'll give people a whole month to contribute and then we'll provide the instructions and how to contribute, and we'll probably do some, like you know getting the word out on that as well, so um we'll probably be talking a little bit about it in the meetings in October forward to that all right. Well, thank you for attending have a good week.

B

Thank you, bye.

A

Bye, okay, now I'd like to talk about two additional papers, and so uh uh these are in complexity and morphogenesis and phenotypic evolution. It's just kind of a broad category, but you'll see where we're going with this. So the first paper is uh it's a scientific reports paper from this year and the title is robust, morphogenesis by chaotic Dynamics, so it combines morphogenesis and some of the models we use for morphogenesis and looking at the dynamical systems aspect of that.

A

So a lot of the papers that we've seen talk about sort of different, stating morphogenesis in terms of differential equations or in terms of cellular automata and then showing that they produce these dynamics that are periodic. So we have things like rule 20 or rule 30 in cellular automata, where we have striping in our implementation of differential equations. So we have striping, that's caused by different gradients, morphogenetic gradients, and so what they're saying here is that you can produce robust morphogenetic patterns or morphogenesis using chaotic Dynamics, so that the abstract reads.

A

This research illustrates that complex dynamics of Gene products, which are, of course, uh the expression of genes and they're, going to show that you know this idea. The concept of amorphogen, which is basically just a generic vehicle for different types of gene expression in the production of gene expression, enable the creation of any prescribed cellular differentiation pattern.

A

So this means that you can take this basic model of coordinated gene expression and you can produce different patterns that are very robust. They describe a lot of different modes of activity. They describe a lot of different regions in that space.

A

This complex Dynamics can take the form of chaotic, stochastic or noisy chaotic, Dynamics, so they're in in chaos and in dynamical systems. We have different regimes for chaos. We have different regimes for things that are on the edge of Chaos and we have different regimes for noisy behaviors.

A

So there they'll talk about these different types of Dynamics. Based on this outcome. In previous research, it is established that a generic open chemical reactor can generate an exceptionally large number of different cellular patterns.

A

The mechanism of pattern generation is robust under probation and is based on a combination of Turing machines during instability and Wooper. Gradients. Wolpert, of course, is Lewis Wilbert, and this is the idea of great morphogenetic Radiance trying and stabilities or chemical structures that exhibit instability is a different critical points.

A

Entering machines are, of course, the idea of running a tape, and you know seeing if it halts at a certain point, so this kind of integrates a lot of different things. um I've not heard of Turing machines really being used in morphogenesis I've, heard of Turing, instabilities and gradients, of course, but not bringing that together. Turing machines in.

D

A

We'll see what they're getting at here, these results can help us explain the formidable adaptive capacities of biochemical systems, so you know there's this idea.

C

That morphogenesis is.

A

Self-Organized and it generates it's generated not by a centralized top-down mechanism by a by a set of bottom-up processes.

A

So, for example, you can have Regional differences, you can have differences between cells between parts of tissues, even within cells and I. Think in in the last two years, we've seen a lot of examples of this in our readings in the group. So we've had, you know different types of single cell morphogenesis tissue, morphogenesis, organismal, morphogenesis and we've seen all these examples of that space is very important. Spatial differentiation is very important and you end up with these sort of gradients.

A

You see things like striping or spotting or different types of bands that are formed in different patterns. Take a pattern formation. You get regularities, how do those things emerge and the ideas that you have? No Central controller? You have these cells or these Gene circuits that work in a bottom-up fashion. They coordinate themselves through different constraints and they produce these patterns.

A

Now my mom be tempted to say: well, we want to have a top-down mechanism that we should have a supervisor that coordinates all these spatial and homogeneities and mechanistic in homogeneities, but that's not actually the way it works. It does work the robotimal process, and so this is a very good example of emergence and they're going to talk about. You know sort of they're going to talk about the field in there and talk about this dynamical systems model.

A

So uh you know, one idea is the idea of Lewis Wilbur's, positional information, and so positional information is where you have a cell and it's in a location in an organism where it's in a location relative to other cells in.

D

That positional information can.

A

Be used as a differential mechanism to describe you know the formation of patterns, so why don't we draw this out so positional information is this idea of where you have a bunch of cells? And why don't we just do this? Why don't we just have like random cluster of cells? So I could show you what the positional information is here.

A

So each of these cells have a position. Let's say it's in an X Y plane.

A

And so we can describe each of these cells in an X, Y plane, and so you know, there's a difference here between y there's, a difference here between X there's, a difference here between X and there's a difference here between y.

A

So you have like these different. You can take differential measures of any two cells, a pairwise relationship where sometimes you can take this in groups. You know clusters of cells that share kind of a common set of positions versus others and so forth. So this is just a random collection of cells. Now, as we know, organisms are not random collections of cells, often oftentimes they're regionalized, so that you get very early and morphogenic very early in embryogenesis. You get polar polarity, so you get PP polarity.

A

Sometimes you get left right polarity, and so we get this. uh You know this. These different parts of the embryos, so the cells in these different parts of the embryo are going to be sort of have a positional identity, and so some of these cells might be an anteriorable.

A

Some of these cells might be in the posterior pole. Some of these cells might be specialized for the right hand, side and the left hand side oftentimes. You know you'll have stereo pairs of cells, so you have. You know, that's something that we see a lot and see elegans, for example, where they have a common identity, and then those identities diverge over time as specialization becomes more implicit or or explicit over development.

A

Positional information, then, is just the sort of the location of that cell. So, for example, we can take this embryo and we can drill stripes across it, and we can say each stripe is defined by a set of cells and each non-strictor is defined by a set of so right there. The positional information is is that they either exist or don't exist, Within, These, stripes and again this doesn't need to be imposed from the top down. This could just be an interval sort of thing so in in in drosophila.

A

You have a whole system for um striping and segmentation, a set of genes that that encode for these kind of setting up these gradients.

A

But if we have these gradients defined, then we can actually have each cell within the striper, outside the stripe and Express different factors that can actually uh might sort of diffuse out of the cell over space and then those signals being oppositional come into contact and form a gradient, so that gradient becomes like the boundary between those two chemical gradients emanating from each cell, and so they set up these boundaries and you get striping, and so all you need to do is specify within a cell what its General location is and again you don't need to specify it.

A

It's specified by nature, but also you know if you were building an artificial system, you might specify this at the level of the cells, not at the level of the Home Group. The point here is: is that it's a product of Gene regulatory networks within the cells? It's not a product that we impose from a top from the top down, and we can characterize this in different ways. We can characterize this with using something like a cellular or automata, which is a grid of cells that interact.

A

So if we want to have a a gradient, we can have a cell here that influences its neighboring cells cell here that influences the neighboring cells and you can see very easily where the cell grid will. There will be a boundary here, for example, between these two cells, because their influences is kind of meeting at that point, and so it's they're suppressing each other and you get into boundary there or.

D

You could have.

A

Something like um a set of differential equations.

A

So you know we might look at a set of differential equations as a way to characterize. You know these coupled relationships. So that's that's where we're going with this.

A

You know what they're doing is they're plugging in dynamical systems into this, so they talked about trying and as a reaction diffusion mechanism, which I kind of showed that you have diffusion of factors out of cells, that the reactions at certain points and those certain points of interaction form boundaries in both conceptual models, possessional information and reaction, diffusion, an organism is represented as a pattern consisting of different cells.

A

So this is just where you have cells in different parts of the organism that end up forming different patterns.

A

Both Turing and wolpert's approaches the cinemorphogens- and these are these theoretical constructs, which are these non-descript factors they could be. You know a certain like in drosophila. We know the specific genes, but in other organisms doesn't really matter what the genes are, what the mech, what the molecules are as long as you have these, they they have a certain action. Morphogens can change cell States.

A

Modern research has shown that French flag and reaction diffusion models do not capture all details of real biological patterning systems, but neither of these captural details are real biological patterning systems. So there's there's some some work to try to reconcile this and, of course, people have talked about the shortcomings of these models, so we're not quite at a full level of understanding of all this, at least through the account the current models um in this paper, uh they're inspired by the ideas of Brenner and Jacob uh Jacob's ideas about evolutionary tinkering.

A

So this is references five and six. We show that a combination of reaction, diffusion and uh positional information mechanisms give a universal robust pattern mechanism. So you can combine these two to produce a universal, robust pattern mechanism. Runner emphasized the importance of turing's ideas in the context of his famous machine. So this is uh where we're taking turning morphogenesis and mapping onto the turing machine. It's interesting, a turing came up with these ideas, but never really pulled them together during his life, and so this is I think people were kind of thinking.

A

Well, you know pattern formation and morphogenesis is the kind of computation and so they're making this connection.

A

The turing machine is a mathematical model of computation that defines an abstract machine that manipulates symbols on a tape According to some rules. The machine stops I'm going to deserve to prescribed terminal state and its turn. Jacob introduced the concept of tinkering evolution uses all available means. So this is where you're talking about this tape. It has different symbols on it.

A

The machine stops in this case when it arrives at a prescribed terminal state, and so the idea is that you can Tinker with that state. You can Tinker with the tape, so the tape has all the stage stored on it, and you just run the tape until you hit the right combination, we show that the Dynamics of a generic open chemical reactor, so they're using an open chemical reactor is capable to generate an exponentially large number of different attractor points which may be chaotic periodic or of a steady state nature.

A

So what they're doing is they're saying that these symbols are these symbolic states on the tape can be represented by attractor points in a dynamical system, the chaotic or periodic attractors can occur from a reaction. Diffusion model is not gradient like and not monotone, so this actually speaks to this reaction. Diffusion model not being gradient-like.

A

We also don't want to monotone. So this is kind of interesting because it means that you can have chaotic patterns form so.

C

A

Linear gradient, there's, no, it's not! You know kind of a smear they're different in other ways. You can have like kind of local states that are differentiated uh the results hold under certain mathematical conditions, which admit a transparent chemical interpretation for two component systems in in this case, one of the reagents is neither an activator or an inhibitor for the other reagent reagent diffusion coefficients are sharply different in the space dimension of the system is greater than or equal to two okay. So that's the first result.

A

So we have a two component system. One of the reagents is neither an activator nor an inhibitor for the other reagents. So you have two reagents you're kind of independent, very agent diffusion coefficients are sharply different, so that means that they're, the regions are different in some way and the space dimension of the system is greater than or equal to two. So the space is larger than the number of reagents.

A

D

Second result is.

A

That such open chemical systems can be used as sort of a turing machine which can produce an exponentially large number of different cellular differentiation programs. So this is where they Define cdps or cellular differentiation programs.

A

This is a program by which morphogen concentrations produce cell patterns unfolding in space and time. So, for example, you can have gastrulation patterns or some mitogenesis, which generates periodical structures, and the number are possible cellular differentiation programs increases exponentially in the system, size.

A

So this is kind of you can think of cellular differentiation programs, maybe like an instruction set, but maybe that's not the right way to view them. But that's that's what they're going with here?

A

The proposed cell pattern formation mechanism is robust against fluctuations and perturbations. So a combination of two of turing's basic ideas, machines and instability of chemical systems.

A

It's a general results which sharply extend turns in stability model to a computational realm.

A

The combined touring walpart model makes it possible to obtain evolving over time patterns that are much more complex than zebra stripes, so they give the example zebra stripes it's a striping example that we saw here in the slide. I mean you can see this on a zebra where you have black and white stripes.

A

They give that example, but what they're saying is they can generate different types of morphogenesis and different patterns that are much more complex than zebra streams. um So they're, not linear. The mechanism of the attractor and pattern generation can be sketched as followed. One can show that a generic open chemical system is capable to generate a number of chaotic or periodic attractors. If it involves appropriate spatial heterogeneity, so this is the these. Are these heterogeneities we talked about before uh you know where things are different across space. You have to have.

A

You know a great deal of spatial heterogeneity for this to work. These and homogeneities work as activators and excite the formation of sharp sharply localized segments like drosophila segmentation, which we mentioned and their specific genes that are involved in drosophila segmentation uh segments correspond to local Maxima of slow, reagent concentration.

A

So this is where I guess the speed of the rage and concentration or a different type of region concentration. You get these local Maxima and you get segments there. This doesn't necessarily map to a biological system. This is just a model that they've developed, so it doesn't necessarily map to drosophila segmentation, but does explain drosophone segmentation, which is interesting. If we know a mechanism is involved in something, and then we have another model, that's as well that maybe not as isn't based in the biology so much but can replicate it.

A

Then it's kind of interesting, because we don't know you know- maybe maybe the solution that happened in nature is one possible solution to this. Maybe there are other ways to do that. So it's kind of an interesting lesson, sort of in evolution segments correspond to local Maxima of silver region concentration. These segments interact in a long-range manner by the fast diffusing component.

A

All these mechanisms reminds Wonder described in the manuscript, but we are capable to prove that we could lead that these lead to chaos. The chaotic Dynamics, as well as to generate complicated cell patterns and space and time, and we show them other results from these works. Okay.

A

So this maybe we'll go to some figures figure. One here is this picture illustrates the concept of a developmental program on the x-axis, which is here here, is a one-dimensional organism composed of various cells that form layered patterns of different time points. Different types of cells are shown in different colors, so the red versus the blue versus the green versus the black, the top row of cells, emerges as the last moment. Tn. The previous rule appears as T equals TN minus one and the bottom row Rises at the initial time opened.

A

So this is uh just kind of describing this part of the x-axis. Is this one-dimensional organism uh they form these layered patterns? These layered patterns are these cell types, these cell types? You can see that they're stratified across the organism uh and then they arise at different times so I'm assuming and they don't label this. But the y-axis is time. So you get these emerging over time.

A

So you basically get this layered structure. It kind of emerges over time. The further down the organism. You go, the more tissues you get and that unfolds over time. So you can see the morphogenesis proceeds a piece we refer to this two-dimensional spatio temporal pattern consisting of cells of different discrete types as a cellular developmental program CDP.

A

So you can see that figure two shows the generalized French flag model, so the French flag is three colors and you have to achieve these three color schemes, there's layers.

A

So this is not just the the two you know the on and off striping of this example, but this is actually three different states kind of layer like like so and.

D

A

Is this image shows a generalized French flag model? A salt pattern can be considered as a string of the alphabet, red, blue or green. So it's this alphabet of size. Three.

A

This pattern can be produced by chaotic Dynamics as follows: be a vector morphoges, morphogenetic concentrations which lie in a morphogene concentration species script, U, suppose that the space is split up into three subdomains script: U sub, J J one two three If U is included in this script. You sub one! uh Well, you know it's one of these numbers, one of these in the index morphogen concentrations induce differentiation in a red cell, and if it's included in uh script, U2 or script u3, then one has a blue and green cell respectively.

A

So, basically you have these different uh I guess these different concentration spaces and, if they're included in one or the other, so you have this index one two three If U is included in one of these. uh Then it's going to express a different color, and so when you Dynamics, is governed by a chaotic system, there exists moments in time such that you state enters uh for uh script UJ. At a certain time. Moments there exists T and then change in t such that the state changes laying the corresponding domain and then this domain.

A

So basically what you're getting is you're getting this. This three uh State system that's being generated over time, so you can see this is these are the same axis as in figure one. So this is the on the x-axis is the one-dimensional organism on the y-axis is time so as time unfolds, you get a more complex organism, but you get the striping where you get uh State one: the blue State, the Green State, the red State, the Green State, the Green State, the blue State, the blue state they're just generated in this way.

A

So over time you get this larger organism with more layers of different. You only get three tissue types, but the tissue types are variable and they don't all draw from the same. They don't draw the same tissue type. Sometimes you get paired like green, green and blue blue red red, but those it's it's a it's a stochastic process as to what the layers are. So, let's go into the discussion here, so they consider phenotype morphogenesis and it's robustness.

A

The pattern formation, robustness problem initiated by seminal papers, 15 and 16. I've been in an intense discussion for the last few decades. uh A number of different approaches have been proposed, chopped proteins and other molecular chaperones, methylation micrornas emerging to embedded mechanisms which are based on Gene regulatory networks and other non-linearies in development. These are all things that have been proposed to this. So what they're saying here is that we can model this using dynamical systems theory and bringing this idea of computation into the sphere. So it's kind of interesting.

A

The general idea can be formulated as follows: timing, instead of tuning, we can take any dynamical system which generates complicated Turing pattern. Dynamics are turning complete Dynamics and we know a number of such systems and then, if, if a Target pattern is predetermined, this Dynamics these Dynamics generates the pattern.

A

The mechanism of pattern formation is robust, with respect to noise and other perturbations, and if it goes up for several stages, Each of which, upon reaching a certain terminal state, ends up with a stop signal. Some such multi-stage morphogenesis, where each stage finishes with the stop signal, is robust with respect to perturbations the idea that Evolution can use any means to produce effective structures which is called evolutionary. Tinkering was proposed by Jacob in six and we're going to go down to the references in a minute here.

A

We assert that practically any dynamical system with complicated large time Behavior can be used to generate complicated patterns. Now.

D

A

Interesting assertion, because we know that cellular automata can do this to to an extent and.

D

A

Don't mention that at all this is just kind of going through turning machines and linking that to trigraphogenesis, but we know that uh cellular automata can do a similar thing where you can generate patterns uh given a sufficient set of rule rule set, and so you know we see sort of how computation intersects here.

A

um The other thing is that I'm not quite sure that uh you know I, don't I, don't know they didn't give a great demonstration to this, um and you know the stop signal is interesting. This idea of stopping the tape the proper symbol at the molecular level. Of course we have stop codons, and a lot of that is, you know very different from what we're talking about here, but this idea of being able to stop uh reading a gene at a certain point producing an outcome.

A

um It's it's kind of there's kind of a connection there I, don't know what that connection is, but just kind of popped into my head, so um so note that the experimental data show that many genes exhibit an oscillatory Behavior. So a lot of times, gene expression is a the output of Gene regulatory networks as oscillatory.

A

It could be over a day. It could be over a month and patterns, usually don't form instantaneously. They form over time, and so these genes can exhibit oscillatory Behavior. Sometimes a cell cycle sometimes over longer time scales, and so we have to consider that as part of the circuit as well, maybe the acceleratory behavior is being used to generate patterns.

A

So, for example, if there's a pattern, that's generated in space- and it's also- you know, solitary with respect to time- that the spatial pattern will be discontinuous, because the oscillations are the peaks of the oscillations are reached at different time points and there's an interval in between where you don't have Peak Behavior. So you end up with these stripes.

A

uh You may also have something like you know where you can't form a pattern any at any. You know over time. Sometimes you can't form patterns and then sometimes you can form patterns. So, for example, if a gene turns on maybe every night where it's photo period dependent in some way, then you know you can have pattern formation during the day, but development proceeds all through the clock. So you know at night you won't have better information stuff.

C

A

That that that's actually kind of interesting model describing the chaotic behavior of transcription factors is proposed in the research where it has shown that such a chaotic Behavior could be useful for survival and leaves the formation a heterogeneous cell populations.

A

Hey yeah, so they used so a solitary Gene Dynamics to describe somedogenesis, but uh you know: that's that's something. I think that needs to be thought about a bit more I'm thinking, primarily in terms of the outputs of Gene regulatory networks, which we've talked about.

A

These results show that there are chemical systems which would be considered in the white of analogy of universal Turing machines, so these are Universal computational machines like cellular automata, but they're, saying that these are Universal Turing machines.

A

The universal turing machine can make all computations, which can be done by other Turing machines and so utms or Universal. Turing machines generate all possible string outputs well when we vary their input. In our case, we have fixed up to a few parameters to adjust spatially extended systems which, depending on our initial data, generate different developmental patterns.

A

So we have these spatially extended systems, they generate different developmental patterns and so we're able to generate. We have these Universal generators of spatial temporal patterns and a few genes are sufficient to encode these it's natural. To expect that such chemical media could appear as a result of biological evolution all right.

A

So, let's look at the reference list. There are a couple papers here: Brenner Sydney Brenner's life's code script, which was published in 2012. That's one reference for this, and this is basically arguing that you know it gets. Development is some sort of like code or there's some code of development. This is interesting. We talked about the differentiation code in this group.

A

There's also positional information, but basically there's this idea that there's a some sort of plan- and maybe that's true- maybe that's not but um yeah, it's worth thinking about this is uh Jacob Evolution and tinkering from 1977..

A

This is Francois Jacob, and this is another paper where you you know they talk about tinkering and evolution, which means that you have a basic body plan that emerges and then it's a matter of refining that body plan shaping in different ways through development um yeah. So you know there are a lot of uh papers in here talking about pottington, Siegel and Birdman, which is uh Bergman, which is waddington's, cannibalization Revisited. That's an interesting paper um Levy and Siegel Network hubs buffer environmental variation in yeast.

A

So this is a paper where they talk about Network, biology and environmental variation being sort of having a buffer, or you know the structure of gene expression networks uh being able to buffer environmental variation, um and then this one here have a siegelman computation beyond the Turing limit. This isn't science from 1995..

A

D

A

Are a lot of different? You know types of things that sort of they're drawing from here um from complexity Theory from biology okay. So the second paper I want to talk about is this paper from the bio archive? And it's also very recent, and this talks about morongioni, like tissue flows, so I don't know we'll find out what those are. But these are you know, tissue migrations that are occurring in these flows and this certain type of tissue flow enhance a symmetry breaking in embryonic organoids.

A

So a lot of uh they're, all French authors, this paper I, think were a bunch of Russian authors so, um and so the abstract reads during during early development of multicellular animals, so self-organized to set up the body axis, such as a primary head to tail axis, based on what which the waiter the body plan is defined.

A

Okay. So this is where the body axes are set up from these primary axes, and this is what defines the body plan and they're saying that cell self-organized, so they don't have this top-down plan which they're using they're, drawing from like local interaction in a spatially heterogeneous.

A

Several signaling pathways are known to control body access formation and this again, these Gene regulatory networks that are controlling the formation of these axes since axial control from Individual cells, depending on their position in that Mass, and then you get this sort of order which is based on these axial cues and axial structure.

A

Here we show, however, that tissue mechanics plays an important role during this process, so they're entering tissue mechanics into this. This is different from the last paper where they talked about turning work for Genesis and reaction, diffusion, which is basically the diffusion of molecules across the space.

A

Also positional information, which is just position. We don't care about mechanics. So why don't we talk about this a little bit and draw this out? So again we have our positional information.

A

We have our spatial heterogeneity. So, let's, let's do this? Let's draw an embryo again.

A

And we have our cells and they have their own positional information right.

A

Yes, I have we have the spatial spatial homogeneity, so the cell here is going to be different from a cell here. This.

D

Is an oppositional.

A

Information, this is just the nature of the cells they're different in different ways, and so.

A

Why don't we leave we'll spatial one homogeneity.

A

I'm not getting in and out there we'll just keep that in mind.

D

Of course, we have.

A

These local interactions, these reaction, diffusion interactions that occur between cells and they set up a boundary if they kind of interact. Otherwise they just kind of have this gradient around each cell.

A

The gradients intersect at some point and that's enough to form pattern formation. But what they're saying is that they're mechanical forces, so these cells just don't sit in the embryo just in one place and then form these patterns a lot of times they move around this embryo in different patterns, so sometimes they move collectively. Sometimes they move at random, but they're always migrating around from their origin point. So a cell divides. It produces two daughter cells. Those daughter cells will migrate now as they migrate.

A

They also produce forces they're also subject to flows and forces of other cells moving, so other cells are moving around collectively, sometimes they encounter jamming, which means that there's such a density of cells that they stop moving. So these are all we're subject to all of this, so our pattern formation is Complicated by motion and by this sort of uh these patterns of motion. They didn't talk about this in the last paper, which is kind of interesting, because you know if you're talking about uh pattern formation is a dynamical system.

A

This is probably something you should probably bring up. um In any case, we have flows, we get migration.

A

Which are flows and jamming and, of course, interactional forces, so these cells as they're moving against one another away from one another or you know, even as they're interacting chemically they're also they also have forces that interact.

A

So you know if it's a matter of like cell squeezing in between other cells, where cells being blockaded from a certain place by a jam of cells or something that's, the interactional forces cells produced forces they move. Sometimes they are in different parts of the embryonic substrate, so they move differently these sorts of things.

A

So this is where they're coming from here, uh so tissue mechanics play an important rule. We focus on the emergence of a primary axis, an initially spherical Aggregates of the mouse embryonic stem cell or an aggregate amount of embryonic stem cells and the mouse embryo, so the mouse embryo starts as a massive aggregate aggregated spherical cells.

A

This is uh this is an artificial system that they've made up. It mirrors events in the early Mouse embryo, so they basically have this spherical aggregative stem cells. They you know, this is something you see in the mountain Rio, but this is not the mouse embryo. So you almost it's almost like in one of these embryoid models where you're trying to approximate certain things about early development.

A

These Aggregates break rotational symmetry to establish an axial organization with domains of different expression profiles, for example the transcription Factor T bra and the adhesion molecule molecule e kit here. So this is adhesion to a substrate for the expression of transcription factors that are very specific to axial organization, combining quantitative microscopy and physical modeling.

A

We identify large-scale tissue flows with a recirculation component and demonstrate that they significantly contribute to symmetry breaking. We show that the recirculating flows are explained by a difference in tissue, surface tension across domains, kindergioni flows well.

C

A

Flows are a specific type of flow with specific type of Dynamics. So we'll talk about what that is, which we further confirmed by aggregate Fusion experiments. Our work highlights that body access formation is not only driven by biochemical processes, but that it can occur that can also be Amplified by tissue forms.

A

So this is where we have this pattern: formation reaction, diffusion model. We have a other types of things going on.

D

And then we have.

A

These flows that actually amplify some of the pattern formation and the pattern formation then also contributes to symmetry, which we've talked about in some of our work. But if, as you, if you need a refresher or I'll kind of explain this to you a little bit, um so you have an embryo which is sort of symmetrical here exactly symmetrical.

A

You have Two Poles an anterior and a posterior pool, so this is a planet which they're symmetrical that some oftentimes, even when you get this poll start to be established as in C elegans, the posterior pull may be slightly smaller than the anterior pool, so you don't really have quicker symmetry there. Ap access is asymmetrical at that point, and.

D

A

As this, the both of these pools, the cells start to divide and differentiate, you get further differentiation of the symmetry, so you get symmetry breaking events and.

D

A

The next term, by word for physics, where people talk about symmetry breaking where you have a symmetrical thing and then suddenly there's an asymmetric shift. So that's what they call symmetry breaking the Symmetry is broken at some point, and usually you can think of this as a tree. So you can think of this as like.

A

A symmetry. That's broken here is a pure symmetry up here. The Symmetry is broken here. It's further broken here and that's further broken here and.

C

A

In a in an embryo would be where, if you had differentiation between an anterior and a posterior Pole, and then that posterior pool got smaller and maybe it differentiated into a germline and maybe some tissues or specific to the posterior end of the organism and the anterior end may be formed a nervous system, some other structures, but what actually leads to more symmetry breaking is if some of these cells start to migrate collectively into different locations, which means that those cells are, you know not like contributing to this sort of equal mass, and so you get this differentiation.

A

You get the Symmetry breaking leads to tissue formation, leads to other things. So any one germ cell, that you take any one stem cell, is it divides. It differentiates it forms a structure differentiates. So if you go to the back to that, Mouse uh massive amount, stem cells, you just have a bunch of individual cells that look basically the same.

A

They do basically the same job they're, not in the same position in the mass, because there's some spatial variation in terms of position, but there's really no variation in terms of signaling or in terms of physics. Well, Pac-Man. Now is this: this massive Solstice differentiate and the cells divide. You start to get things like this or you get differences in position, differences in kind of location and maybe like migration, outward into little modules that are like little arms coming out and.

D

So you get a bunch.

A

Of different types of variation here in terms of how these things differentiate, things flow outward where they flow. You know around the cell outside of the cell Mass, so the cell Mass, because less of a spherical mass and more of sort of an irregular drooping, but you also get these clusters of cells that begin to have different functions, and so these aren't, you know uniform you get chemical signaling between them.

A

We've got positional information that defines where they are in this case, positional information on the left side in a positional information is less important because it's just kind of like you know, you're either. You know you're either in the middle of the pack or outside on the outside. There may be some differentiation there.

A

That's minimal in this case, if you're on the outside, it's quite a difference between being on the inside and being on the outside and B uh migrations complicate this because they move things around.

A

D

What they're saying is this.

A

Amplifies the existing processes.

A

Which is perfectly fine statement to make, and it's probably true, but but you know, this is the kind of thing people use these terms that are very, um they kind of come up with terms that sound good. But then you know what is the amplification mechanism?

A

um Can we can measure how much it amplifies things? That's another story.

A

Our work highlights that body access formation is not only driven by biochemical processes, but it can also be Amplified by tissue flows. We expect this to operate in a number of contexts, so we know that there are a small number of conserved morphogens, including wind, BMP, active and nodal and fgf in in animals that control the emergence of body axes. These signals participate in tissue, patterning and Trigger differential behaviors, which ultimately partition an isotropic group of cells into different territories.

A

So this is what I mean by the Symmetry breaking that these events actively participate in this Progressive symmetry breaking. Where things become more and more differentiated, they talk about tissue, mechanics and flows being crucial for these stages of symmetry breaking. So it's not just genes, it's not just chemical signaling. It's not just its position, but it's migration. It's like mixing things. Ironically, this mixing uh contributes to symmetry breaking and it's because it doesn't mix randomly it mixes in a certain way that moves things to certain places.

A

We know that tissue, mechanics and flows can contribute to embryonic axis formation avians, but we don't know the exact rules in mammals, so here's an example of what they're doing here. They look at the SE tissue, Collective flows. They have this.

A

This Vector map here, where they're showing the directional flow things, are moving in different, basically in a coordinated Direction amongst a bunch of cells, they're moving posteriorly a lot of the time, so this is contributing to symmetry breaking the.

D

Others live Imaging.

A

At this sort of migration, and then you have expression, brachi or you get here and so uh then also snail. This is live Imaging. This is immunostane just showing the different patterns of how things, so this is actually gastroint symmetric breaking. So this is in a gastroloid.

A

This is kind of showing this process averaged over one hour. They find that tissue flows substantially contribute to gastroyed symmetry breaking. We. Actually we see that you have these different sort of stages of things, so you have reaction, diffusion part of their model, so you have cell differentiation and then you have diffusive motion. So there's this reaction, diffusion aspect where things move around, but it's diffusive, it's according to say, like Brownian motion, they don't move like in a coordinated manner.

A

They just kind of wobble around they have this positional identity, but it's there's diffusion of things, and then you have this advection stage, so you have diffusive motion which is basically Brownian, and then you have this Collective motion which is non-brownian super linear, and so the cells move around according to matter. This is what they call advection, so things are moving around. It's it's becoming well mixed, but because these cells are, um you know, have different identities and they have maybe a memory of this identity.

A

It's going to contribute to symmetry breaks. Ultimately, so you have Collective motions, as we saw in figure one you move around together and then you have growth, which is cell division and apoptosis. So some cells die some cells proliferate and that's how you get the progressive symmetry breaking if you think about the cumulative change of polarization, which is something they measure quantitatively and looking at hours after aggregation of these cells in this Mass. Just remember the Mast Cell Mass.

A

You see that you can look at different regimes here and if you only consider one Contra contribution of a single Factor at a time you get this graph here to E. So two e shows that you have uh hours after aggregation from 76 to 92 hours. Cumulative change in polarization going up is the as the y-axis goes up, and you see that the total it goes up over time.

A

76 hours is partly visible, but at 92 hours becomes predominant and actually after 80 hours, reaction, diffusion seems to be a major contributor to this advection. As another contributor three-dimensional effects and growth are not really that effect. Much of an effect tissue flow is also exhibit a dominating recirculating component.

A

So that means that there is a lot of fluctuation there's this recirculation of cells. To this end, we Define a disk that encompasses most of the aggregate projection and express the coarse grain tissue velocity in polar form. So they put this in polar coordinates. We then decomposable spatial functions into different modes again using polar coordinates and they look at the circulation here. So this is analysis of tissue flows during polarization. These are coarse bearing tissue flows analyzed in a circular disk, and this shows kind of what they look like.

A

So there's this recirculation mode, where they kind of recirculate back, so you can see that they move in insert at the same pattern. These recirculating flows can be driven by pattern, induced tension, so in understanding the tension and how patterns form forces in the cell. So you have no pattern and then you have the onset of a pattern and.

C

That pattern, then, has.

A

Consequences on the embryo on the cells, so cells can be there induced tensions that can force cells to migrate after pattern formation. So.

C

A

Not just a one-time thing: it's not just you form a pattern and that's it so then they talked finally about the morangoni effect. So we expect we expect that marangoni effect driven recirculating tissue flows operate in many organoid systems, especially systems form for employee potent cells. So in this case we tested stem cells, but there are other types of chlorine potent cells we can use well, such marangoni flows can be artificially induced at what they mean by marangoni effector. These recirculating flows that we see up here.

A

So you can see this very clearly here where you have experiments where you show this recirculating flow and then simulation, you can reproduce it uh well. Such Marin marangoni flows can be artificially induced. They can also be driven internally. The emergence of such flows requires only the existence of distinct cell populations, which will typically display different surface tensions, thus providing a general mechanism for amplification of symmetry braking. For instance, recirculating flows have been identified during Mouse heart morphogenesis or during the creation of the first cardiac Crescent during gastrulation.

A

So this forms a feedback loop that we can observe in potentially many systems. So that's all for.

D

Today hope.

A

You learned something.