DevoWorm Lab Meetings, 14 Aug 2023

Previous Meeting

Next Meeting

⏯

youtube image

►

From YouTube: DevoWorm (2023, Meeting #30): DevoLearn, DevoGraph, Minimal Cells, Molecular Order, Differentiation

Description

DevoLearn/DevoGraph updates. Topological analysis pipeline and persistence diagrams from static data. Resources for synthetic biology and minimal cell evolution (evolving from a minimal state). A theory of molecular order? The central dogma and theories of the cell. The steps towards differentiation, understanding asymmetry in lineage trees, and cell-cell interactions. Attendees: Sushmanth Reddy Mereddy, Bradly Alicea, Susan Crawford-Young, Jesse Parent, Richard Gordon, and Himanshu Chougule.

A

Oh here he is hello.

B

A

Are you hi, hi I didn't miss anything last week.

C

You didn't oh yeah.

A

C

Didn't have a meeting.

A

I I was um yeah, I was on Lake of the Woods okay.

C

A

Explaining that that's I, like this, between Canada and the US, it sort of splits the lake in the middle.

C

Oh yeah, yeah yeah.

A

Yeah so he probably sailed into the U.S a little bit yeah.

D

Well, Bradley yeah can I tell my update. If you don't mind.

C

D

uh These many days, I, was having so many books related the implementation of diamond is actually so. Finally, I cleared out, all the bugs and model was able to train and the loss function is also less than the last 118 models. Last debonairs models last one was about lost, mean loss was around 1.2 around that, but now low sponges are around 0.99 and it is furtherly decreasing. It is at epox zero. Only I haven't tried this because of some GPU issues. Maybe by tomorrow, I will train it.

D

The code is working, fine and I have completely worked on and I pushed the code. Actually with this all code, I worked on this week is complete I just pushed this code and tune it back and right now the good thing is Sam model is working and download is also working. I just need to review I. Think yeah right, I have learned so many lessons. While implementing these all code writing code is not important, but in a structure way is more important than I.

D

Think so so I have implemented everything here, but they are not in structure way. So I was thinking to make it structure to create tools, different reports for development model and Sam, also because Sam is mainly used for cell segmenting. The owner will be used for acquiring the cell volume cell area, etcetera and cell positions. So why not keep it two different repos for if we mix two projects in one repo, it is quite confusing.

D

Actually I thought combining it into one into one but yeah it's confusing, so I was thinking to buy, who different reports has two different projects: okay and I'll. Do the open, One, Foundation thing yeah.

A

D

The update would be pretty simple: I have worked on the technical issues this week, completely the loss functions and implementing the Sam model uh have an hour approached Sam model of training, but the model was able to train and learn now, and the loss function is also less as we expected, and maybe the accuracy will be also good too I just need to add some augmentation to the data. That's it and make a clear documentation about these two projects separately.

D

I see where right now, we code and all this thing so I want to make two different reports for both of them and add some pictures and do a clear documentation would be useful for.

C

Super yeah that would be great um yeah. You said something about the code being like harder than you think, or what would it? What were you referring to there.

D

Actually, the port I've done the core part very quickly. I mean records uh solve all the bugs with help of mayuk yeah, but the situation is uh there is a mistake from myself: uh I haven't selected the code in a perfect ring, and this is the thing creating problems. Now the code was not in a structured way in a two different reports. If you see Bradley this, this repo contains of two different models. One is devonet, one is cell, segmented, okay,.

D

Creating so many problems, so I was thinking to create two different repos more like separate and sell, sounds operate segment emitting models from creating clear documentation to both options. I mean the. If you see here, the model is working actually I thought it says. Segment in anything model will not work at that, but yeah here it is working, I mean haven't these are this? Is not the error?

D

I just created a keyboard, interrupt very, very stock, but I know right now the model was able to learn and it is trying to get the launch in as we expected then past level and models. So my next couple of weeks would be the work of organizing this all code in a correct way and make a clear documentation and Bradley. I have and I'm really sorry that I haven't attained last week. Meeting I was asking you and I was admitted in the hospital, and the hospital has some.

C

Sick rules that we don't need to with the laptop right.

C

Oh that's, fine, yeah, yeah,.

D

Just make a clear documentation and the paper trap to the whole of this thing and I have found that Springer has a free publication kind of thing for us thinking to write a springer in applying intelligence. Okay, so yeah documentation, I have started. The documentary. I haven't worked on the documentation yet, but I will start from this week and make a clear documentation.

D

Next couple of weeks would be writing a clear paper about like how the model is working and comparison between the models and Etc right now the model was able to learn and it is training also, so there won't be any problem.

C

All right uh give me a little bit more detail about the data augmentation or rethinking there.

D

The documentation mainly I, was thinking, but right now we have around 4950.

D

So I was thinking to it flip it flip the data and in different angles and pair it in different ways and applying different filters, like it's a gray scale, image right. Why don't we convert into its RGB image and actually pythonch library has different data. Augmentation techniques such thing to apply those only to fire. This documentation, I just told these examples of clipping it rotating in a different angle and feeding it. This would.

B

D

The model accuracy and or make real different uh test cases for the model to learn not on a single test is right away image the model the image is flipped in a different way, and if it is budget, maybe model was not able to learn it. So I was trying to do the data. Augmentation apply different filters to it like a green filter, hello filter and just a edge detection.

D

If we apply these filters it will work, but definitely fine, I, think so. Right now we have the best. uh I mean bus law best loss because it is less than one and model was able to learn. If you see here, there is a difference from zero here it is decreasing. But if you see here 255- but here you can see that it is decreasing from that, so loss is decreasing and model. Here it indicates that model was able to was able to learn the whole features from the data.

D

If we add more data augmentation to it, Etc, even the loss function may go below 0.5 of my idea. Okay,.

C

Yeah, that's great.

D

Yeah next couple of weeks, Bradley I will completely work on the paper and the documentation part, because every project might insisted me to make the documentation a correct frame. We are having a meetings or sat down. Sunday we are discussing about it uh yeah. He told to make a documentation clearly and the EU and warned me that he will fail me if I don't do the documentation is a problem. You need to learn.

D

Actually I was updating him everything so yeah, okay, this is uh this is the complete work I have done this week, adding the loss functions to the owner and trying the sell, segmenter sorry model, removing all bugs, so both of the models were able to learn. So next couple of weeks would be a writing a paper and comparing it over different models like unit.

D

That would be my total implementation for a couple of weeks, making a clear documentation and then rewriting a clear question.

C

Yeah yeah I think that's a good plan. I uh I went over some uh resources on writing documentation if you're familiar with the meetings and I think you went to one that we run on open source uh on Fridays and in the or uh orthogonal lab slack I have a channel I, think you're, also in that channel Google summer of code and I put some resources in that channel. So those are good resources for talking about how to structure, documentation and uh and I know, if amancio's in that channel too.

C

So there's a lot of stuff on how to structure documentation, how to write uh clearly, of course, but also how to what to include- and you know how to make it accessible to people. So that's I mean basically what we want to do is have you know like a readme file, of course, in the repo, but then also we want to have you know, maybe some more detailed documentation for uh you know you want to have something for the general public.

C

If you just stop at the repo and you you're interested in it, you want to give people enough information to say this is something I might want to fork, or you know star or do something like that then later on, you know you, you get a little bit more deeply into the software and you say: well, how does this work or how does that work? And so that's where the more in-depth documentation comes into play, and then you can link that to to like Publications.

C

So we've talked about the devil, worm white paper or the pre-print that can be, like you know, more detailed uh structure where we have you know maybe some examples, some use cases and then that links to the documentation and saying see this for more. So it's basically just you know you want to make a very accessible version. You want to make a more in-depth version.

C

Then you want to have your Publications linked to that, where you really give a lot of detail, but you also don't want to you can't give like every single you can't get into like code examples and things like that in a paper, because it bogs down the paper. You know, so you want to kind of push that into the deep documentation as well. So it's yeah I.

D

Will comment out the whole course because someone wants to see the code if they don't understand it could create a problem. So I will add the documentation to the code on so Computing every line. That would be the better thing, I think so, because right now, devoland doesn't.

D

Try to comment every line, what I wrote, and maybe that would be useful when someone reads the code um go through the code, what is happening in manual yeah? That's where my updates, okay, yeah.

C

All right, yeah and yeah commenting is always important because it helps people walk through it without having to flip back to the documentation. So.

D

But it got extended in the main official page. Thank you for that. Okay,.

C

You're welcome yeah all right thanks schwantz for the update um amancio. Are you there.

B

A

B

That, if they got out before implementing it, is uh so basically compared to our other systems. Our topology is dynamic, uh so uh basically our cell diagrams, like the cell nucleus, segmentation, type uh images and membrane images, so we can find find out the topology of cell dependence of Time video frames.

B

uh So we can start from taking extracting frames and then segmenting, which is the first uh of a project which is similar for both me and sushant. And then we can take off line Point cloud of that and then do TDA on it. So it will be a static topology if done this, but the hard part is to make a time now, uh because the time release said uh image and it's because it's video frame sense moving all the time. So in this case our approach would be similar to PG.

B

So the first step would be to do the implementation and to figure out these two questions like how GNN can be applied to cell development analysis and how it can perform with TD on multi cellular system and another approach. I had thought of was to do percent persistent homology over Point cloud, data uh of which is what my example is built upon today. uh But it's it's an image example because I wanted to start with an image and then further go through Point clouds.

B

uh Before that, I would like to explain the like topological layer that was presented in the 2022 paper.

B

This is a graph like it's basically a set of node attributes, and you heard okay, so this is basically a graphics. uh This would be. Basically, the daughter and mother cells are uh in case of, like our cell elegant a CL against data, and this is basically a set of nodes which is in the Rd space like uh each uh topology has like a Zero Dimensional space.

B

One dimensional space, two dimensional space Etc and uh the first step to do is uh to we have to do- is to create this node attributes and convert it to a node map which basically Maps each node uh to some value, uh and it's uh basically like a subset of the entire thing. And after doing that, we get like okay different views for graph G, because we are applying k y for K naught values uh to a network map. So we can. We are getting like different views of the same graph.

B

After doing this, uh this is the main part of the paper is like doing the filtration process, and this is where, like uh uh filtration process and on the each and every View which gives us a persistent diagram and persistent diagram, is like.

B

Persistent homology is like a built upon like yeah every. If you, the filtration, is doing the.

B

Federation is doing the topological graph competitions and then we are getting a persistent diagram and the aggregation step is basically uh two like incorporated into a neural network, GNN layer. So in the aggregation step, it's how we like uh sends like the graphs, are permutation and variant so, and uh uh it is it's also permutation. uh It is a permutation invariant. We have to do this aggregation stuff and after we do that, really like the topological graph layer and sensor.

B

This is a topological graph layer. Since it's a layer we can apply it on any GNN offer type or theoretically. So we can also filter this a built upon this, uh which was done in the previous resource here, which was basically temporal crafts nutrients, which.

A

B

Had worked on so uh if we can uh code this layer and we just incorporate it, uh so it should work like basically in a Varina GNN as well.

B

uh So this is what I had figured out or that this is one of the approaches we can do and for the example I have proposed like uh uh a function which I also, which I thought would be greater enough uh as a as an example, to show that a TDA works on cell data as well, uh which is basically doing a low star image filtration. So what a low star image filtration is that it allows us to express local minimums and maximums uh if we flip the image like if we make the.

B

If you get the negative of the image and what this does is, it gives us per times as well as saddle points.

B

And these points are basically a good way to stabilize the image and give us the critical points which we can look upon, uh and uh this is basically a example like uh which, like origin, block and uh NC reference spaces and uh doing the same thing uh on this test, image will give us a Zero, Dimensional persistence, diagram where the the x-axis shows the paths since then, the this is the dead cells, and this at point five.

B

You have a like a threshold layer which tells us like uh it's where the critical point is acting up. So we can convert this entire thing in a Cell Biology image example, and when we convert this, we have to use the grayskin image, but we always already have a green screen image in our data sets.

B

We have used so I've worked on both memory, interior cell membrane data, as well as cell nucleus data, to see how it changes accordingly, and in this case, what we wanted to give is like a local Max of persistence with each cell and uh just let's see uh over how it is so basically I printed the shape and the values of the original remains.

B

And here this is the uh plot of the segmented image of a new nucleus data and, after that, I have done the same for uh uh our membrane image and I've gotten this uh low star image function on original image. Like the membrane, then this is the graph we are seeing here. So, as you can see the lifetime of the cell, uh it's neared, uh it's always around near nearer to the zero values and it Peaks.

B

At a pointed value- and this is basically the sensor- grayscale image is from zero to one like uh spaces from 0 to 1. This is how the Oh- this is, what the diagram shows us and, uh as you can see it, uh most of the values are near quite clear at the zero and it's like it starts to go up after uh like around 0.5. uh Oh sorry, not point five, uh it's 0.05, so that could be like a threshold which is what I thought of, uh but to get a good uh values.

B

What we can do is to smooth the image so that our cells that are enable us neighbors to each other will have like similar values.

B

After doing that, we get like a better. uh uh That is what I thought of it will give us a better output. So after doing that, uh we get like the more uh more uh points which are nearer to the zero value, and the lifetime of the cell is also changed. Like first set of the maximum was 0.8 and most were nearly 2.05 and uh like this, but now it is uh a little different.

B

So this is just saying that uh most of the points which are we have we, uh we don't need them to give us the uh well persistent diagram that we actually the topological features that we actually need from it like the maximum uh low star filtration uh points.

B

So after doing this, I've put a threshold as 0.005 and uh also uh Five Point uh I plotted the points which were like above the threshold, and this is the points which are we are getting. As you can see, it's uh nearer to all the points which are like connecting the cells and uh I'm getting a four a different out, a little different output. uh It should be like nearer to the centers of the image like, uh which is what I'm working on right now.

B

uh So this is one of the ways we can take a so this is one of the ways we can uh basically give get. Topological features from a data set of our choice. uh Also, one of the things I'm working on, which would be like uh good for the project, uh at least like in terms of visualization, is paraview.

B

So this is one of the uh tools I've used like during my uh College during my college, for a project and apparently computer. So uh this uh this is like real, and this also has a topological toolkit, which is why I was really interested in it. uh So I was also figuring out like, and this also gives us a lot of topological uh functions and features. So we can also do topological data analysis on this and uh in.

B

And we can then also compare both of them like how so we are getting like uh if we, the topological graph layer, how it's giving us persistence uh data and also we can get persistence. Data from this toolkit and I'm still learning the toolkit and right now uh the data which we have right uh for, for uh which we use for the uh it's not fitting inside the paraview, so I'm figuring that out as well and I'll get back to you with that with that as well.

B

Also uh like this is my entire update and uh my further things would be uh to oh look. uh Look at that, because here record, uh which is basically uh the or the open source repository which was given by the iso or what the topological graphene paper and from this I'll have to uh basically do some similar jianga did for his uh g-soc. That is like to you uh to take this.

B

Library kind of thing so that we can uh do it, I can just speak. The.

B

Plus figure out a way to incorporate it in a Cell phase data set. So this is like the part which I was mostly uh confused on, like how will I implement it. So this is- and uh this is what I've thought of like in the start, which, in the start where I showed that embeddings answer uh this part.

A

B

Oh yeah, so so this is my date, so it was uh uh a little uh all over the place because I've been working on it for a few weeks and I was not able to give a proper update. So I apologize for that, but yeah. This is what happened till now.

C

Yeah, that looks actually pretty good. uh Thank you for the update. Yeah you've made a lot of progress on things and.

D

C

Looks like you, you've worked out kind of a pipeline for for producing something from the data, so that's good, uh so you're basically plotting the the graph yeah you're, plotting the point cloud and then you're plotting it out and then you're transforming that into basically topological features which you could build a graph from. If you wanted to and you're putting that over here, you overlaid that onto an image or I thought I saw that in the last entry in your notebook.

B

uh Yeah and then the image- and there is another example I'm working on, which is to do it on a Time series data. uh But since our image sincerity, the set of this again against Richard likes um relation between uh daughter to mother cells. It was also time series based, but it was not entirely like uh time series because it did not have like specific time stops on it, so to figure out a way to convert it into a away so that it gives us like time, series plots and once I do. That.

B

I can also uh do a similar filtration method on that graph as well and as well as I can import that into the topological toolkit which I showed earlier. But that would be yeah.

C

Yeah I think the time series component would be important to you know, make something where you could just plug in data and actually have the time series data, and then you know build some topological features from that, because you know we've done a lot of things with static images, but uh the time series is important for capturing some of these changes.

B

B

Yeah also given like a uh or website where you can see the entire CL against three branch like from the start to like all the mother daughter cells, uh I, can also try to uh get that and the data set and to.

B

Created I can start working on that.

B

C

Good yeah yeah the interactive uh C elegans lineage trees. What he's talking about so that's yeah.

B

Yeah yeah yeah.

C

So that that actually has all the cells- uh and you know you- can uh it.

B

C

Annotation data too, so it tells you what the cells are and then you know you can use that as as your Baseline. So that's good I, don't know exactly how it exports, but.

C

Was great uh so yeah, that's great and then again uh the same uh things that I talked about with sushma in terms of documentation also apply to your project. But you know you I, guess you're planning on some papers. I saw you had a reference to maybe a target for a paper in your notes.

B

Yes, so I know as well, but uh first I have to figure out the implementation part of it and after that I have like uh so I'm still in my final year. So I have like I'll. Do that as well.

C

Yeah yeah is kind of working on a accelerated track where he wants to put on all these papers, but yeah I mean.

A

B

C

Pressing thing here is, like you know, having good sort of annotation of the code and then having some documentation so.

B

Another thing like right now, I'm just using one note, uh but since it's like uh in uh particular pages and it's kind of maintained and use that as reference and as also I I'll, be creating like reports for my college internship reports as well, because that's one of the mandatory things I have to do uh since it started so. I'll I'll also have that for a reference, so that I can do it like really well I think uh also the oral guys also have a report I think in colleges as well. Foreign internships.

C

B

Yeah, so it's kind of similar that way only so I'll have to do that and uh I can I can show, show it to you as well like what are uh posted over there. All right. Thank.

C

You all right, thank you, yeah great.

B

Thank you for my time. Yeah.

C

uh It looks like dick and Jesse have joined us since we, since I last, uh got a break in here. So welcome uh yeah, hello, uh so yeah. We we that those are our gsoc updates and uh I've been working on the paper. Actually, it was uh Lucas and I were working on the paper uh for the uh differentiation codes, so we're actually doing a lot of work on analysis, getting the analyzes together.

C

So this is where we're doing all the methods and getting the data together and so I talked to Lucas Friday and he's done a lot of work on this uh so far uh and we compared notes so we're getting we're coming along on that and uh yeah I. Don't know if Lucas is going to join us today, but I think we yeah we're pretty close to getting to a point where we can review the document. I don't want to do it today.

C

It's still not quite in the to the point where I can walk through it, but it's it's! It's coming along uh some pretty interesting findings about some of the things that we're looking for. Okay. So, to recap: for those who don't know we're doing this bioinformatics analysis we're looking for sort of Target genes that dick and Natalie mentioned in their book uh that might be involved in in differentiation and putative cell State splitter. So we have the protein list.

C

Lucas has been busy getting protein matches from uh ncbi he's been doing, blast searches, uh getting was looking for a list of uh proteins in their uh homologues and paralogs across four species. Actually we have drosophila C elegans, which is our Target and then Mouse, and so that corresponds to okay yeah. Thank you, I want you. uh So we have C elegans, which is our sort of focal species we have drosophila, which is another example of Mosaic development. We have Mouse, which is regulative development and then East, which has no development.

C

You know so there's no developmental period there and the idea would be those proteins would be, you know, have different counts for those different species and then in C, elegans I'm, looking at some of the genes that code for those proteins. So if you go through the literature, you can find examples and see elegans, it's a wonderful Library.

C

uh They have not only the sequenced draft genome, but they also have a lot of function for different uh genes and you know so their papers where people have found uh genes coding for proteins, yeah so great, but please leave time for rest of us to review which you found yeah. So uh we should have this done by the end of the week or to a point where it's like a draft document where we can go through it. So let me share my screen. I have a couple things to share with you today.

C

um The first I want to talk about are a couple references going back to some of the synthetic biology stuff we were talking about and some of the Earl, maybe even early life stuff we were talking about so the first thing we were talking about was this coding guide to synthetic biology. uh This is something that Nico McCarty put out. Nico McCarty is he does this uh uh this uh site on? He writes this thing called codon, it's a sub stack.

C

That has a lot of interesting content, but this is a nice guide he's put together on synthetic biology and some of the readings that you might need to know if you want to engage in this area, so um I I shared this in uh one of the channels in the slack and uh Morgan who's. Not here today really thought it was a nice reference, and so that's that's good. This, basically just gives you different references to different areas of synthetic biology, so this is going to be a living document.

C

This is something that's going to be updated uh over time. So, uh let's see so, there were a lot of seminal paper links here. uh There's the basis basics of synthetic biology. So during these paper on the foundations for engineering biology, which is way back in 2005 now, but that's on the list, a brief history of synthetic biology which was published about 10 years later and synthetic biology doesn't just go back to 2005. It goes back to a lot of the work.

C

That's been done on building constructs to deliver things like fluorescent proteins or uh you know, Gene, uh Gene, inserts and things like that into cells. So a lot of the stuff with crispr comes out of a long history of uh modifying cells putting genes into cells knocking in genes even knocking out genes. Things like that. uh So that's that talks about that. um Then there's synthetic biology applications come of age. So a lot of this is not just for experiments where you show a difference.

C

If you take a gene out or put a gene in uh these are Gene copies or sometimes they're the gene itself, but you have this. These applications that you want to build. So you know different applications that are useful to agriculture. To Medicine um this one is the cell really a machine. This is in the Journal of theoretical biology. This is interesting because a lot of synthetic biology is predicated on this notion that the cell was a machine and that we can work with it like a machine.

C

We can take Parts out and put parts in and make corresponding changes. So there's always a correspondence between like a parts list and the function. So, that's, of course an assumption that people make, but is it really a fair assumption, and so they talk about that in this paper, um then the rest of these references in this section are, you know, just kind of talking about research tools which change very frequently. So some of these uh you know papers are maybe a little out of date, even though they're only maybe about 10 years old.

C

um You know there's some forward-looking papers, which is now not a well I guess it was. uh It seems like a forward-looking paper, but it's actually a retrospective, the second decade of synthetic biology 2010 to 2020, which was published in 2020. So it kind of talks about that, um and then you know the second wave of synthetic biology for modules to systems where you know, we think of sometimes like genetic systems as modules like Gene Gene regulatory networks, groups of genes that work together to produce some, maybe phenotype or function.

C

But you know Thinking Beyond the module. So if there's a module for some, you know to produce some enzyme, that's great, but that fits into a larger system. So that fits into the system. The the Genome of the cell and the function of the cell, so all these things are sort of you know they have to go for modules to systems and that's this paper talks about.

C

uh Then they have this section of the central dogma. We'll talk about this paper 60 years ago, Francis Crick changed the logic of biology, some things on the central dogma, some references on gene expression and regulation, so these are really for the uninitiated, but some of these are really kind of interesting, like this unified theory of gene expression, which is kind of a reinterpretation of what people are.

C

You know people think about like the relationships between gene expression and proteins, and so a lot of the things that we've been discovering, especially with respect to synthetic biology. Are um you know they have this sort of uh these outstripped the theory that we have for it? So now people are trying to come up with new theories around this. This paper is actually about 20 years old, so I, don't know what the follow-up has been, but we'll talk about this later as well, uh then their biological networks and Mathematics.

C

So this is only a few papers in this area, but uh you know you have things like artificial Gene, regulatory, Networks, mathematical, biology and bio modeling and synthetic biology, modeling and Analysis of Gene regulatory networks and so forth.

C

uh Then there's this graphic notations, so this is kind of getting into systems biology and some of the things that people are doing to represent things uh graphically. So uh you know some of these things involve, like uh you know, Open Standards, there's this synthetic biology, open language, which is a community standard for communicating designs, so in synthetic biology or say, like designing.

C

A gene circuit you're designing how it's supposed to be put together, and you need a unified language to build that it's it's kind of like programming where you have a unified modeling language, but they have a different way of doing this with synthetic biology. uh Then there are these graphical notations that you can use, there's open language, visual, uh which is an open source, graphical, notation and so forth, um then their Technologies, so these Technologies are, are always changing.

C

They're, always new technologies coming out all the time- and you know some of these, though, have stood. The test of time so see DNA sequencing generally, is you know a similar process for the last 40 to 60 years, so sequencing doesn't necessarily change it's just the way in which we do it. uh There are different techniques for doing it. So there's this paper DNA sequencing at 40.. This is by Jason duri.

C

This was a this just kind of talks about some of the different sequencing techniques that have been used in the last 40 years: uh Next Generation, DNA, sequencing methods which you, if you get into next-gen methods, you'll find that there's just this massive proliferation of methods.

C

So this this reference is from 2008 but I think it's it sort of puts those a lot of the methods into uh or order for people so that they can understand what methods do what so some methods, sequence, DNA uh and and do certain things to the say- the amount of Gene product quantify them. There are different ways that you can quantify them: they're different ways that you can sequence the DNA. There are also methods for uh quantifying and sequencing epigenetic uh characteristics. Things like that.

C

So this is a you know, there's kind of a zoo of different methods here, um and it may be out a little bit out of date because there have been a lot of methods since 2008..

C

In any case, it's it's probably the reference to go, uh go into then there's DNA synthesis, so people in synthetic biology need to synthesize DNA synthesizing sequences. So this there are a number of papers here: DNA assembly, where you're assembling, fragments of DNA and putting them together.

C

There are other types of methods of cloning. This is cloning genes, not cloning organisms, but they're different.

B

C

To clone a gene, DNA cloning molecular cloning techniques, isolating cloning and sequencing DNA. There are all these ways that you can do this.

C

This is basically PCR and Associated methods, um there's crispr Cass, which is of course, this method, where you can Target different stretches of DNA and either cut them out or put in new pieces, and so this is important in synthetic biology because you're trying to build these circuits- and you know if you could put something into an existing next to an existing promoter or something it helps- you build a circuit, and so there are different ways you can do this, and this gives you crispr cast is a very popular method for sort of modifying genes, building a new function into Gene circuits and whatnot.

C

So this is a number of different references on this. It's a relatively new uh method, maybe not relative to a lot of molecular biology but like in terms of biology. It's recent! It's it's! You know, since 2015 really um there's base editing, which is editing, DNA bases, there's Prime, editing, there's directed Evolution, which is where you know the try to evolve things towards a goal. uh This is a technique: that's used a lot of times in proteomics uh protein structure, prediction and design, genome synthesis and then applications so there's biomaterials, which is a popular area.

C

There are a number of references here, a number of uh review, articles, biosecurity and bio containment, uh biosensors, which are actually useful for things like arsenic, biosensors uh and other types of cell-based, biosensors, sensors or water contaminants. Things like that. Then there are these cell free systems which are I'm not really familiar too much with cell free systems, but they do have a number of readings here and then Gene circuits.

C

So of course you have to understand Gene circuits, it's sort of like the building block of synthetic biology and they have a number of references on how to build Gene circuits and those all come from the wac operon model, uh which was developed in E coli and so they've been. You know. This is like the basic level of sort of Gene regulation and understanding understanding it and manipulating it. And so, of course this goes beyond the black Opera and you can do a lot of things in different organisms, but in E coli yeah. If.

A

Someone wants to know about the lack of operon, they have uh from North Dakota State University, they have a virtual cell and they have a cartoon of the whole laptop opera on series. Oh.

C

A

It's it's very understandable.

C

Oh great great, that's great, thank you um yeah, so actually in E coli, you can do a lot of really interesting things. You can engineer E coli to see light. You can engineer them to do different things to produce different Gene products and there have been some interesting experiments where people have engineered bacteria uh more generally to do a lot of uh cool sort of like lighting up or spelling out things in a in a dish or whatever.

C

So there are a lot of really interesting uh like demos you can do and interesting things you can build.

C

um So this is not just theoretical people build things all the time now, with uh largely with bacterial genomes, because they're easier to work with, but um you know so then their Gene drives, which are uh this is a paper, a genetic control of mosquitoes. So there are ways you can build.

C

Dean drives that are uh allow you to control things like uh different, these that allow for mosquito mitigation, um and so that there's of course, some things you need to be aware of, like some issues here uh so they're engineered microbial communities, there's metabolic engineering, plant engineering, recoded genomes, synthetic cells, so synthetic cells, that's another really interesting area, and this is uh where they're building cells from different types of very minimal cells and they're able to build sort of the molecular aspects of the cell and plug them in.

C

So this is programmable fusion and differentiation of synthetic minimal cells. This is transmembrane transport in an organic colloidal cell mimics. So these are things that are just containers for the cell or for the molecular stuff, and the thinking is is that if you go back far enough in early life- or you know to some very simple organism, the cell is basically this container for a lot of these molecular reactions, and things like that. So you know this is again an area that is uh being innovated upon all the time.

C

So uh speaking of that last Point, uh there's an article in Quantum magazine, I think it's a recent article on uh on these type of minimal cells and this the title is: even synthetic life forms with a tiny genome can evolve so uh by watching minimal cells regain their the fitness. They've lost. Researchers are testing whether a genome can be too simple to evolve, and so this paper, um let me magnify this a bit more.

C

So this paper uh talks about the system. uh So seven years ago, researchers showed that they could strip cells down to the bearish fundamentals, creating a life form with the smallest genome, but still allowed it to grow and divide in the lab, but in shedding half its genetic load. That minimal cell also lost some of their hardiness and adaptability that Natural Life evolved over billions of years. That left biologists wondering what the reduction might have been a one-way trip.

C

uh So this reduction may have been a one-way trip in pruning the cells down to their Bare Essentials. Have they left cells incapable of evolving because they could not survive a change in even one more Gene?

C

So basically, what they're saying is that if you have this sort of minimal genome, you know- and this is something they do in a lot of uh like uh this- this area of synthetic biology where they they find these minimal genomes and it's usually a bacterial genome where they do these combinatorial nachos, so they basically knock out every Gene until they get you know in the the DNA there allows the organism to survive or it doesn't. If it doesn't, allow it to survive, then it's considered an essential Gene.

C

If it does allow it to survive, it's considered sort of a Superfluous gene or a redundant Gene, and so from that type of experiment you can assemble a minimal genome, and so this is what they're getting at this minimal genome and if you knock a gene out of that minimal genome, it can't really survive a change in a gene, because, theoretically, it's sort of locked into that it needs everything to survive and there's no path forward.

C

um So this is uh dick, says what if vesicle changes, that's volume to near 10 to the 11 times uh constant or zero 10 to the 11 times, consequences opportunities?

C

uh I'm not really sure what that means, but uh near zero. Okay,.

C

So if you it's, if you change the vesicle volume to near zero as you're saying.

C

Yeah I, don't know it's a good question um early Earth spun fast. So this goes back to the stuff we were talking about with early life, so yeah the. If you your day time is very short and your cell is very small. Yeah I, don't know yeah it's a good question, I'm, not sure yeah.

A

10 to the 11th.

C

During the origin of Life yeah.

C

So in this paper they they talk about some of this work on minimal genomes and these minimal cells, so they say minimal cells, immune minimal genome, and so uh now we have proof that even one of the weakest simplest self-replicating organisms on the planet can adapt during just 300 days of Evolution in the lab, the generational equivalent of 40 000 human years.

C

So this is where they're actually testing this time uh so they're, actually looking at like they're doing when you do laboratory Evolution, usually pick uh organisms that have a very fast replication time with DNA or with even with bacteria. You can do this pretty easily, because the replication period is, you know within maybe a day or so or maybe less than a day in 300 days you can actually uh simulate, or you can experience forty thousand human years.

C

If you you were to take like human evolution and compare it with the evolution of this genome 300 days would equal 40, 000 years uh measly minimal cells regain all the fitness they've sacrificed. So this is a team from Indiana University. This is a paper um in nature, evolution of a minimal cell, um and they talk about this. In this paper, uh the researchers found that the cell responded in selection pressures about as well as a tiny bacteria from which they were derived.

C

A second research group at the University of California San Diego, came to a similar conclusion independently in work that had been accepted for publication. So it turns out that life, even such simple wimpy life, is a minimal sell, is much more robust than we thought.

C

So if we have just this minimal set of genes that are essential for the cell to survive, and you start evolving it you put it down the path towards you know, replicating and mutating, and things like that, eventually you'll end up with a more robust uh genome than you started with, and so you can start off from a point of absolutely no redundancy just the essentials, and you can end up with these. This variety that allows you to have genes that are that are sort of uh redundant and other things as well.

C

So you can actually have this very rich genome that you can that aren't all you know all not all the genes are essential, you can throw a rocks at it and it's still going to survive even in a genome where every single Gene serves a purpose and a change would seem be seemingly detrimental. Evolution mold organisms adaptively.

C

So that's basically the article. This is the minimal cell uh from the Craig winter Institute. This is uh sin 3.0. So this is a strain of bacteria built up in 2016. They work with us a lot.

C

um This is a a minimal version of the parasitic bacteria mycoplasma, my guides, so they strip down the Genome of that organism and they have the cell line now that they use- and so this is uh you know this is this is kind of what we're talking about. uh So this is a bacterial genome. This isn't uh uh Madison genome, so medicine.

A

C

Or any kind of linear genome is is much different in a lot of ways, so but I don't know of any experiments. People have run on those. Mostly these have been run on bacterial genomes.

C

um Going back to the uh papers, I promised you before. This is the first one. 60 years ago, Francis Crick changed the logical biology.

C

So this is a nice historical piece on Francis crick's lecture in September 1957, in which he outlined key ideas about Gene function and in particular, what he called the central dogma and these ideas still frame how we understand life, the SSA explores the concepts he developed in the influential lecture, including his prediction that we would study evolution by comparing sequences, so this is sort of the historical base for a lot of the work we talked about in the synthetic biology reference uh page.

C

uh You know talking about these different things and really a lot of this work on molecular biology comes down to what was proposed in this lecture. um So this kind of talk this originally entitled protein synthesis. The title acquired a magisterial introduction on during uh writing up for publication the following year. uh The lecture went far further than the title suggested, as Crick pointed out in the opening paragraph. He also addressed the other Central problems of molecular biology.

C

Those are Gene action and nucleic acid synthesis, so this is kind of what was put together and what was later called the central dogma or the central dogma lecture, for it was here that he first publicly presented this frequently misunderstood concept. So basically it's the idea that things go from dnas to transcripts, to translation, to proteins. uh So things move information moves out from the DNA to the proteins, and so he kind of uh you sort of outline this because before you know, I mean before the 1950s.

C

We didn't really know that DNA existed in the in the double helix structure that we came to know, and so we also needed a theory to put together a lot of the you know DNA and protein and some of the other things in the cell. So this is where this kind of it wasn't really a theory, but it was a framework for thinking about this.

C

It also saw quick outline his view of the nature of life and of genetic information and the source of protein folding, as well as making tubal and spectacularly accurate predictions.

C

The first was that there must exist a small adapter molecule now known as TRNA, that could bring amino acids to the cytoprotein synthesis and then the future scientists would be able to explore the rich evolutionary sources of information by comparing sequence data. So in the first point uh he basically predicted TRNA, which exists in the ribosome, so you get uh transcribed uh DNA.

C

It goes to the ribosome, it gets translated uh or it gets transcribed by TRNA into amino acids that then get built into proteins, and a lot of this happens at the ribosome and there's some fascinating work. That's been done on these TRNA pools in the ribosome and how they can be selected for different codons or how they're biased towards certain types of codons, and this happens in cells uh in all different species. Sometimes it's species specifics. Sometimes it's self-specific bias, but this is you know this is sort of the basis of old Square.

C

uh The second part is that we should be able to explore evolutionary origins by comparing sequence data, or you know, evolutionary sources of information, and so this is the other part that that he talked about so yeah. He talks a lot about the central dogma. He sets up this framework. He builds a model of in the central dogma, is a lot more complicated, Nuance than I just talked about so, uh but they basically outlined all this, so he outlined the sequence, hypothesis and protein synthesis, um so they kind of came up with this idea.

C

That uh DNA is the structure in the double helix, but they also found that there are these chemical bases that form a sequence, and so these sequences that are are transformed into these proteins, and so they were able to link all that together.

C

um In doing so, uh you took these ideas and increased it and found that the data increasingly suggested that RNA was some kind of intermediate between DNA and protein, so we fit RNA into the monor. uh These data refer to ribosomes rather than mRNA, but they figured out the sort of the structure and basic structure of this and developed a scheme to explain the relations between these three classes of biological molecules.

C

In doing so, we had to get to grips with what exactly was a gene and what took place. Ef DNA was used as a template for RNA, so this is again you know putting all these things into into the proper place, and this is the test of time. So this is not just a biochemical problem. This is an abstract thinking problem. This is a model that you have to build. So uh besides a nice article, uh this other paper is the unified theory of gene expression.

C

So, as I said, you know, a lot of these findings in biology are sort of just findings, but you have to put a theoretical net around them now with synthetic biology, you wouldn't think you would need to do that. You think you just need better tools, but in fact you also need better hearings, and so in this case they propose a unified theory of gene expression. This paper is about 20 years old, so it's not a recent paper. I, don't know if this has stood the test. The time or people developed upon this.

C

But let's, let's see uh the human genome has been called the blueprint for life. This master plan is realized through the process of gene expression. Recent progress has revealed that many of the steps in the pathway from Gene sequence, the active protein, are connected, suggesting a unified theory of gene expression. So this is, uh you know, taking a lot of the things of the cell that are going on and kind of casting a theoretical literal.

C

So this is an image of things in the cell. This is a traditional view of gene expression, so gene expression um occurs when you get a transmembrane receptor, something binds to the receptor. There's a signal that's sent downward into the nucleus through the cytoplasm. This is transcription Factor activation those through the nuclear pour into the nucleus. It triggers uh chromatin, which opens up a gene. Then there's this promoter region which gets initiated. Then there's this elongation and terminations.

C

These steps, where the gene is transcribed and then a RNA mRNA product is produced, and so then this is where you get this five Prime three prime orientation uh that mimics what you see in the Gene and how the DNA is laid out. Then you get this capping. You get this processing of the RNA, so you get the secondary structure, mRNA packaging and then it's exported from the nucleus and sent out into the uh cytoplasm.

C

Then there's translation, which occurs at the ribosome, which we just talked about where the MRNA is translated into amino acids through these trnas sort of taking the templates and stamping them out into amino acids, and then you know matching what's on the MRNA template and then you get proteins that emerge from those amino acids and then protein, folding and then proteins are active and that's where you get so. This is the standard sort of model based from the data and what we know now, unified theory of gene expression, would kind of bring this together.

C

So um they talk about some of these steps. These are just this is kind of like a process model, and then you know kind of arriving at a theory.

C

You would come down to I think it's not two, but three.

C

So this kind of goes through a lot of this process, um and you know you kind of consider some of these steps as different. You know with different insights, so order DNA, binding and chromatin remodeling. It promoters insights into a chicken and egg scenario. So this kind of talks about how uh the requirement for chromatin decompaction for transcription Factor binding DNA appears to create a chicken and egg scenario. These factors in these local chromogene remodeling that their interaction DNA requires prior chromatin decompaction.

C

So how then the transcription factors access the DNA in the first place, and so you can consider this. You can do this experimentally, but you can also do this theoretically and think about different examples and how they play out. I'm.

B

C

Really sure this qualifies as sort of a unified theory never really comfortable when people throw that around, because it's just kind of like bringing all the pieces together whether they actually came up with a a sound theory, is another question.

C

um But this kind of you know at this point in history. This was when there was a lot of yeah excitement about molecular biology, so it wasn't quite that we had like um I mean the human genome will just been sequenced, so you know we're entering a period that will revolutionize the study of gene expression. uh We know enough about the individual steps of the gene expression Pathway, to begin to understand the connections between them and the processes that regulate them. The traditional view of the cell is a bag of molecules.

C

Separated into compartments by membranes is beginning to be challenged. I guess that was the whole point of this framing. This is a theoretical question, um and so you know we, we know the parts we know the process and we need to build the theory. Maybe maybe the theory of a cell is too ambitious if we need to dial that back to like a theory of the nucleus or something and then move from there. So now, I would like to talk about a few papers related to the molecular biology of differentiation.

C

So this is relevant to our paper and differentiation codes that we're working on, including one paper that answers the question dick asked last week or two weeks ago about asymmetric divisions. So, let's get started.

C

The first paper is on e-cad Heron and this paper is titled, an instructive role for C elegans ekit Heron in translating cell contact cues into cortical polarity, and so this paper talks about sub contacts providing spatial cues that polarize early embryos and epithelial cells. So basically, when you have early differentiation in the embryo, you need spatial cues and when we talk about spatial cues, we're talking about things like anterior posterior polarity, we're talking about polarity during cell divisions. Things like that now cell contacts are the contacts between cells, provide these spatial cues.

C

So, for example, when you have a cell that has polarity. So if you have, let's, let's take the two cell example from C elegans, okay, so these will eventually get onto their respective subtle lineages.

C

Now, if you notice that these two cells have contacts here, so there's a polarity here that allows it to sort of maintain this like kitchen. This is what they're talking about these cell cell contacts. The polarity is anterior posterior, and these are the cell contacts, and so he could Heron is a protein that allows contact induced polarity in many cells.

C

So, as this two cell starts to form more cells in each part of the for.

A

C

Anterior and the posterior and get more cells that sort of emerge from these two sublineages, their cell cell contacts here that are important, Anika adherent is important in maintaining these cell cell contacts and in maintaining the polarity across the embryo.

C

This is one reason why these embeddings that hamanchi was talking about are important, because we can build these kind of models and look at the things that mediate. These connections between cells.

C

However, it is debated whether you can't hear in functions instructively as a spatial, cue or permissively, by ensuring adequate adhesion so that cells can sense of their contact signals.

C

So is it a spatial cue or is it just a signal of sort of the touch touching of the membranes of the different cells? Adhesion?

C

That's an important point, because if you have it's a if it's a spatial cue, then there should be some long-range function to it as well, and certainly should be that if maybe you took a cell out if there was a mutant phenotype, for example, where there was a cell missing that there would be maybe a gap in the embryo or you know the cells migrate a lot after they divide, so maybe their migration. If you took up you get Heron either migration would be screwed up. Something like that. That's what they're getting at here.

C

They don't know whether it's a spatial, cue or some sort of adhesive function in C, elegans contacts, polarized early embryonic cells by recruiting row, Gap pack, one which is a another protein to the adjacent cortex, including par protein asymmetry. So you have a symmetry of how proteins are expressed here. We show ekit here in hmr1, which is dispensable for adhesion functions together with Alpha Caton and hmb and hmp1 the p120 katin and Jack one previously uncharacterized Linker pick one, and this binds pack one and recruitative contacts.

C

So you have a series of proteins, a protein Cascade that needs to be in place to realize this sort of uh cell cell contact, um so the real as a function of each adherent in its context, misallocating the hmr1 intracellular domain that contact free surfaces.

C

So this is where we have this hmr1, that's mislocalized, and so this is um in a contact-free surface, so it's just expressed somewhere else uh draws pack, one of these sites and depolarizes cells, demonstrating an instructive role for hmr1 and polarization.

C

uh So hmr1 is an interesting protein and Associated Gene I'm going to be talking about more on the paper. I I just recognize the name as something that's going to be important for uh regulating some of the developmental genes, but in any case, we'll come back to that later. Our findings identify an e-kid here in mediated pathway. The translates cell contacts into cortical polarity by directly recruiting a symmetry breaking factor to the adjacent cortex foreign, so.

A

C

Is an interesting kind of paper that talks about some of these Pathways uh ecad hair and mediates this pathway? It involves a lot of proteins being in place and when they're not in place, we don't have this function for each adherent.

C

So one of the things we're trying to do in our differentiation code paper is talk about each head. Hearing is sort of one of these proteins that is necessary or that should show up as a product of different cell differentiation, and indeed these cell cell contacts are an important part of cell differentiation, because when these cells divide and they migrate, they need to be in contact with one another, but they also need spatial cues to move them around the cell.

C

What their findings are here is that they sort of have this spatial specificity, but they also have this sort of sensing aspect. So you have this complex, Cascade of proteins that need to be in place.

C

If you move the protein expression to another place on the cell, it doesn't operate as as needed, so it has to be spatially localized.

C

You can see that, like in in cells, they need to have this basically the spatial coordination amongst themselves and it's based on the expression of these proteins, where they're localized on the membrane.

C

So, for example, if this this cell here started to migrate in this direction towards the midline, then it would come into contact with new cells and it would need to make contacts with those cells and that would need to be regulated in a different location than if it's making contact with the cells further anteriorly to the to the anterior end of the embryo, so migration, of course, these this differential expression of all these proteins will play a role in facilitating this migration.

C

So their answer here is a little bit of both in the original question, and so they do some experiments here. These are basic experiments where they're looking at the different things that are involved in this mechanism for each adherin, they look at moving. You know expressing it in different places.

C

These are some microscopy images, so this is pack one structure function, analysis.

C

This is the pack one Locus here um and they show the uh the ph and GAP domains the position of xn6 nonsense mutation, so they will get it for different mutations. They look at this uh the stain here, so this is for pack one uh these stains. You can see along the edge of the membrane. You have expression for pack one.

C

This is of course important in this ecad hearing pathway, um and you have pack one here in in green fluorescence. You can see it as well.

C

And so the results are at the packet. One n-terminal domain mediate cell contact globalization, so we have cell contacts, but it's localized cell contacts, a homophilic adhesion protein hmr1, contributes to pac-1 localization p120 Caden and Jack one and Alpha K and nhmp one recruit pack one to hmr1. So we have this Cascade where things are recruiting each other.

C

And then you have hmr1 gfp localization pack, one localization, so you can see it's very similar.

C

And then this is sort of a diagram of what's going on of the cell cell contact so where the cells joined together. This is a cell cell contact. This is where you have hmr1. You can adherin, so you have two of these proteins in each cell and they're, aligning the Jack one and p120 cotton- and this is next to this uh ecad in this cell. Hmp2 beta cat is next to the kid here and in this cell and then hmp17 Alpha catonin is below the beta Kingdom.

C

So this is all happening within one of the cells and the other cell. It's also happening, and then the hmr-1e cad herein are the ones that come into contact between the cells, so that maintains the contact between cells and then within each cell. You have this Cascade of other proteins that are sort of joined to it.

C

It's interesting because in our list we had EK and here and we had beta Caton and- and we had uh I can't remember if there was any other I, think those are the two that are really kind of they're kind of joined at the hip anyways functionally. If you read through the literature, there are other ways in which they're connected, and so this.

B

C

This is an interesting sort of context to this.

C

C

So that's the ecad hearing paper uh now I want to talk about another paper where it's more General Paper about sulfate specification in the C elegans embryo. So this is about sulfate specification, as opposed to sort of cell cell contacts and some of these cell-some interactions.

C

So this is for Morris Maduro um and the abstract read spell cell specification requires that particular subsets of cells, adapt unique expression patterns that ultimately Define the fates of their descendants.

C

So this is where cells are specified for um by either gene expression patterns, and you can identify those using uh bioinformatics techniques. Usually it's the upregulation of certain genes and we kind of think of those as a unique fingerprint or signature of a certain cell type. Now you know you could critique that method and say well, that's only one aspect of it. Maybe there's a lot of overlap with different cell types and indeed people compare different profiles and sometimes.

B

C

Are unique sometimes they share genes that are upreculated, but in any case this is a requirement of cell specification that some genes are upregulated there maybe have a a very specific functionality to a certain cell type, of course, are going to be things like housekeeping genes that are up regulated.

A

C

These expression patterns sort of prime, the cell for the fates of their descendants, so.

B

C

Of course, we know that the lineage tree is deterministic. You go from the top down to the terminal, we differentiated cells and all those cells have to be in place. They also sort of are precursors to that terminal. We differentiated cell. So we know, for example, that part of the lineage tree, for example, will produce germ cells or part of it will produce neurons or muscle, and so in those precursor cells and those founder cells and some of their descendants.

C

You'll get an elevated gene expression pattern of these Fates early on before the cells actually show evidence of that fate. So that's the kind of thing we're looking for, and this is also true of mammalian cells, where you have different precursor types.

C

You get these up regulation signatures in sort of the precursors to a certain cell type, so you have neural precursors, for example, or muscle precursors that aren't exactly the terminal fate, or at least the Fate that we recognize them as, but they do express some of the things that will later go into making that identity and so on. C elegans cell space fate specification involves the combinatorial action of multiple signals that produce activation of a small number of blastomere specification factors.

C

So this is where we go back to the blastomere and we talk about what kinds of things are specified uh factors just meaning the genes that are expressed. So we just we just pick a number of genes that are regulated, and these are. uh This is activation. Activation of these last mirror specification factors actually sort of facilitate these. These signals to sort of drive forward the differentiation process.

C

These initiate expression of Gene regulatory networks that drive development forward, leading to activation of tissue specification factors. So we have blastomer specification factors and tissue specification factors, it's interesting because it's yellowgens, we think of it as a deterministic, an age tree, so it basically unfolds the same way in every organism. Unless you have a mutant, in which case, maybe a cell is missing, or you know it very rarely gets mispecified.

C

um But when you do you know you don't get a normal phenotype, and so this is an interesting thing, because it's it's saying there are two steps to this. There's blastomere specification, which kind of puts it on the road towards differentiation or keeps it on the road towards differentiation. It facilitates a lot of signaling between cells and signaling within the cell, and then these tissue specification factors which are sort of you know what tissue does it belong to?

C

What tissue is it going to belong to it, primes it from being contacts where its neighbors are of the same tissue type or that is part of a larger tissue with other cells participating in this review. The C elegans embryo is considered as a model system for the studies of specification.

C

So it's not really, like you know. A mammalian system where we have sort of you know cell will come into a fate, but it can be driven away from that fate. So if you take a cell, that's maybe a muscle cell and you put it into uh bone or if you take a bone Zone, you put it in a muscle cell. It could convert into a bone cell or a muscle cell.

C

So if you convert into the tissue that it's a part of, if you do a transplant experiment, where you take a cell one type, you put it into another type of tissue. Now it's an interesting point because it doesn't always work that way. So in mammals, you have to have some level of or of pluripotency, in order to get that conversion to be complete. uh If, if you get a cell, that's really firmly differentiated, it may actually not convert fully or may undergo apoptosis, so even in in the everoplastic mammalian models.

C

You see this uh specification that's hard to break and C elegans. It's a nice system, because you don't have this sort of plasticity. You do have cell spell cell specification once the cell is responsible or even before the cells reach their terminal differentiation they're in this sort of mode, where they're going to be they're, priming to be the cell type, and so nevertheless, you can see the mechanism.

C

The techniques used to study sulfate in the species and the themes that have emerged are described. So this is from developmental Dynamics. This is from a from the year 2010, so I'm not really sure. If there's been a lot of innovation in this area since then, but this basically lays out a lot of like how to study this.

C

So you know they talk about uh metazone embryonic development. Embryonic cells must ultimately assign Fates to individual cells and tissues. When a cell has been specified for a particular fate, it is capable of generating those tissues if cultured away from the remainder of the embryo. So basically, if you take a cell out of context, if you take it out of its embryo, if it's been specified for a particular fate, it can generate tissues, it can generate, say muscle cells if it's been primed for the muscle fate.

C

So you see this in cell cultures, where you have cells, you may expose it to muscle factors or it might be pretty far down the pathway and you put it in a dish and you can get muscle cells. So you know a lot of cell differentiation is priming by by its uh molecular environment, it's micro environment as well as just kind of this process this internal process. But if you you know it depends on how far along the way it's primed.

C

So if it's been primed for, say a muscle fate, it will raise your muscle fade if it's not interfered with be any other type of Fate. So if you put a muscle cell, something has been primed to be a muscle cell into a culture of neurons. It probably will convert to some sort of neuron, but maybe not.

C

um It really depends on how far down that path of commitment that muscle cell is gone um as only organic cells are derived from zygote by mitosis. Cells must acquire differences in gene expression over time in most, animal embryos is achieved through a combination of inherited maternal determinants.

C

So there are a number of maternal determinants that that are inherited and c elegans and mammals and invertebrates, um and so you have uh in most animal embryos is achieved through a combination of inherited maternal determinants. Cell cell interactions, which we've talked about in the context of ecad Heron, but also other interactions such as juxtapine signaling and Peregrine, signaling positional information, which means that the cell kind of has additional information about where it is in the embryo.

C

So, for example, if it's in a location where muscle should be or if it's surrounded by muscle-like cells, it will be more likely to become a muscle cell and factors segregated within a lineage from a mother's cell to its descendants. So this is where you're in a lineage in the C elegans lineage tree.

C

It's a deterministic tree, meaning that if the mother cell is you know a certain type where it comes from a certain founder, so the daughter cell will become is restricted to those types, and this is true of a lot of different cell types. Where you get this fate restriction, where you know the daughter cells can't just become another type of cell together. So, for example, if you're in the one of the muscle making lineages, which you can actually map out on a lineage tree, you're not going to suddenly have the cell become a germ cell.

C

It has to become what it's uh maternal cell and maybe what its founder cell allows for in modern, developmental biology. The description of Gene activities that results in specification constitutes a gene, regulatory, Network or grn, and so this is a grn sent to Davidson and Irwin. So this is the type of grn where, if you've ever seen a Davidson uh style grn, it's basically a circuit diagram of different genes and their products and how they interact.

C

So um these are kind of uh you know, they're hard to implement computationally, but they're really descriptive of the biology and how these things I, don't know if they have any pictures of them here. But this is the tight kind of year in because there are many types of grns that people have proposed. So these are Gene regulatory networks that basically link genes together by one gene regulating another gene or one gene producing a product that regulates another Gene.

C

So this is. uh Although animal embryos have evolved different ways of specifying very early embryonic cells, the properties of grns are similar across many systems.

C

Hence the study of cell specification and model systems can be used to exploit the advantages that system of that system to reveal new mechanistic insights into General, grm properties, and so this is one of the things we're interested in or is Gene regulatory Networks uh and we've talked about them in the group.

C

uh In the past uh it's been a while, since we've talked about them, but I think it's always worth talking about them, because they're very useful, they're, very Central to developmental biology, so C elegans is good for looking at sort of this generational aspect looking at embryogenesis because you can get a you have a short generation time of about three days. So you can generate a lot of embryos in three days, um and so you can actually get a nice sample of things that you can look for.

C

Things like you know, lineage trees and you can look at like developmental gene expression, so it originally salston uh mapped up the lineage tree and he was able to uh demonstrate that c. Elegans really doesn't show very much variability from animal to animal s, what they call nearly invariant. So there's some plasticity in the lineage trans, mostly post hatch or post embryonic.

C

So there's so many you just give a really nice model of uh sort of non-variability.

C

um So this is an example of lineage tree. You see these sub lineages, where they produce different outputs, so the a b lineage produces epidermis, neurons, pharynx muscle, Ms produces pharynx muscle, he produces cells for the gut, C produces muscle and epidermis. He produces muscle and P produces germline, so notice that a lot of these sub lineages. These are what they call founder cells.

C

So you don't see the founder cells with an A B, but they're founder cells here, there's MSE, C and P those founders of their eight founder cells and those founder cells, kind of Define. What the subset of or restricts what tissues are going to be made from that sub lineage.

C

So this is kind of like in uh you know, faint restriction and vertebrates where you have sort of this. This process of where you go from Cody potency to pluripotency to certain types of precursor cells that can produce certain types of cells once you're at the precursor cell stage. You know you need to use some sort of engineering technique to get yourself out of that precursor stage, but even so you know there's this sort of restriction of Fate that occurs over time in a lot of these systems.

C

So in this case we have cells in the zygote, which is totipotent to some of these founder cells, which are pluripotent, but actually the lineages that are going to emerge. The daughter cells are going to emerge are only going to be restricted to certain types of cells and in fact, in the adult, you have this nomenclature of specific cells that come from this hel lineage tree. So.

B

C

Isn't a lot of plasticity here, but.

B

C

This fate restriction property notice that their muscle gets produced in a lot of these sub lineages, but some lineages are really highly specified. So e, for example, is highly specified for gut cells only which is the stomach, and some of the intestine D is highly specif specified for muscle and so they're different types of muscle that these sublinear inches produce Ms is basically the pharynx and muscles. So this is like the area where the C elegans is ingesting things and the muscles surrounding it.

C

P4 is the germline, so that's highly specified and then with an A B. You get a more of a mix of things, but it's largely epidermis, neurons, pharynx and muscle. So all these things are produced, they're they're, restricted in their Fates, but they're producing, like you know, um uh you know, sometimes they're only producing one type of tissue, sometimes they're, producing multiple types of tissue. It's an interesting kind of a side, but uh it's interesting how this is structured, because um you know some of this.

C

Of course your C elegans lineage tree is organized from anterior to posterior. So you know some of the stuff going on the posterior end of the worm is maybe more special specialized than the interior end. Now that of course, kind of absurd, because the bulk of your connectome is in the anterior portion of the worm. But you know those neurons. Of course this neuron category is, is you know? It includes a lot of diversity, so I didn't want to say that, like you know, this is just a bucket of neurons.

C

There are a lot of different types of neurons and, of course, when you get down into the lineage tree, your specialized types of neurons- and you know, when you get down several levels of sublineage, you get specific types of neurons that emerge and so that that's that's, basically how this works, um and then this. This is an example of kind of this process from the two cell, the four cell, the eight cell, that what they call this comma stage, which is where you have this.

C

It looks like a comma and the gut is forming at this point, and then you get this so it's a fold in the embryo and then you finally get the gut inside you get the pharynx, the muscle on the outside. So you have these outer layers of muscle, which are all these sub linear edges and then the germline which they don't show, which is kind of like right here and it's just the eggs and the and the vulva. So this is how this works in CL again.

C

So it's a very nice elegant system for this type of thing um and then so. This is the cell specification Gene list, so this kind of talks about these different sub lineages and some of the proteins that are associated with them. So you know you get for the germline. You get this Pi one, the ccch zinc finger.

C

This is the peel lineage. This is germline pal one cuddle, C and D Med, one two data Ms and E, so you're, just taking things out of the literature and finding connections to the these are all cell specification genes. So these are all genes that are active when that cell is being specified into a certain uh down a certain uh it's being restricted to the certain fate. So this is these are some of the uh proteins involved in blastomer identity.

C

These are some of the proteins involved in tissue identity. These are some of the proteins involved in maternal blastomere identity, so there's a maternal and a zygotic component to blastomere Identity, and then you have the tissue identity, which is, of course, this later step, um and so here's the analysis of gene expression and perturbation in normal development. So you get this gene expression that occurs normally.

C

You can also perturb the embryo to modify the gene expression to make sure that you know you understand the mechanism, and so this shows some examples of what they're doing here um and it just vindicates the model that they're working from so another thing they do is this forward and reverse genetics, so these are complementary approaches, so there are different ways: you can do this different methods, in addition to embryo manipulation, cell cell interactions can also be studied.

C

Genetically C elegans possesses a hermaphrodite mode of reproduction, which is mostly how it reproduces there's some sexual reproduction, but it's limited. This is great: we've facilitated the recovery of mutations and developmentally important genes. The first direct of regulators be identified that affected early on embryonic specification. Events were recovered and forward metagenic screens that looked for maternal effects in embryonic lethal phenotypes, exhibiting a deficit of tissues of one type and in excess of another.

C

So you take do these genetic screens where you screen for mutantations, they call them mutagenesis screens, you look for embryonic lethal mutations, so that means that they produce dead embryos. And then you make the connections. So you can actually do these metagenesis screens as a form of forward genetics and then so you can do these kind of screens followed by mapping and molecular cloning.

C

This is been a good way to identify these regulators screens like this, have recovered glp-1, which includes a notch like receptor, important various symmetries in the a b lineage apx1, which includes a Delta like ligand, that mediates glp-1, dependent cell cell interaction between P2 and app, which is we saw a P1, and a b, ABP and P2 are descendants of those two cells. So this is in the I think the four cell stage, where the this Alters the fate of ABP, and so these interactions are altering the fates of different cells, given a mutation.

C

So what they're doing is they're finding all these mutations that affect fate? Basically, that are important for setting up the Fate restriction. So they've had a number of examples here. Regulators can be identified by their expression in an early blastomere lineage pal 1, which encodes a caudible. Protein was first identified by zygotic mutants later found to play a central role in specification of sublinear, just C and D the somatic descendants of P2, following the discovery that pal 1 protein is found in the early P2 and EMS lineages. The same thing holds true for those.

C

Similarly, tbx35 was identified as a candidate Ms factor for the MS lineage because of its expression in the early Ms lineage. So all these things we're trying to link them we're trying to implicate them in this fate restriction in these Sublime ages, and then you can use things like RNA interference uh and knowledge of the genome sequence. So we know we have a draft genome sequence.

C

It was actually done before we completed a little bit before the human genome in 1998, and there's been an update to this uh genome sequence in 2019, which refined it a bit. But you basically have to have knowledge of your genome, your whole genome and you have to have a map so that you can do our own interference. So you can Target certain regions of a gene to knock its expression down and remove its expression and then see if it's if it produces a phenotype, that's viable. So.

A

C

You know forward, genetics is where you're actually looking at mutations that have occurred. Reverse genetics is where you're inducing some sort of change to the genome or to the phenotype and you're looking at the effect, and so you know there are a number of methods you can use for these sorts of things you use transgene. Reporters, of course, to um find you know, get a signal that represents the presence or absence of that Gene or protein being expressed, and so these are all things that you know you can use for this thing.

C

um I do find this important, fascinating paper, because I think there's a lot of things we can learn from it uh in terms of so looking at this differentiation code, so we'll be coming back to that paper, perhaps in the in the differentiation code paper or later on in another meeting, but I'd like to finish up with this paper, and this is actually something that uh did.

C

It was a point dick raised in one of the previous meetings about asymmetric division events and, of course these are important in looking at the differentiation tree, because we're interested in asymmetries and we're interested in differentiation- and so this is the title of this paper- is asymmetric. Division events promote variability in cell cycle duration in animal cells and E coli, which is a bacteria, so asymmetric division.

C

Events are interesting because they will sometimes produce one sort of dead cell or you know, one dead cell and one live cell or sometimes they'll produce cells of different sizes. So there's basically, is an asymmetry being introduced into that cell division.

C

If it's a cell division, that's you know where there's one dead cell and one live cell, then the dead cell will be. You know there won't be a pair that you know there will only be one sister, so so that has implications for a number of things.

C

Also, if you have a difference in size, and sometimes there is a significant difference in size between cells during a division that.

A

C

Be you know have an effect on the later organism, because the mother, the daughter cells, will eventually become other cells and they'll either make a contribution, their daughter cells that compounds over time or if it's, a larger size cell. That's size differential compounds over time.

C

So it's very important for say like an embryonic development and E coli, it's a little bit different, because basically it's a tonal reproduction, but still you have this issue of you know its effect on the colony or its effect on its descendants or whatever so asymmetric cell division is a major mechanism generating cell diversity.

C

A cell cycle duration varies among the cells in mammalian tissue cell culture cells. We asked whether their division asymmetry contributes to this variability.

C

We identify among sibling cells and outlier using hierarchical clustering on cell cycle, durations of granddaughter cells obtained by lineage tracking of single histone 2B labeled mdcks, remarkably, divisions involving outlier cells are not uniformly distributed in lineages. So this means that you have these asymmetric cell divisions they're not uniformly distributed in lineages.

C

They appear to emerge from asymmetric divisions taking place at non-stochastic levels, so these are uh I, guess they're targeted in different lineages in different places.

C

These asymmetric divisions are sort of not randomly distributed across the lineage tree, uh apparent cell influences with 95 confidence and 0.5 error, then unequal partitioning of the cell cycle duration in its two progenies. So this is where, like I said, these things compound, so the progenies carry on these sort of the effects of this asymmetric division. So when you have an asymmetric division, say one daughter cell dies and the other lives.

C

Those daughters also produce progeny, of course, that not only live but carry on. You know maternal things from the Mother cell, maternal factors or the complement of chromosomes that were segregated into that Mother cell all those divisions ago, and they get carried down into the progeny.

C

There's also, you know, uh there's also information. That's lost by that dead cell same thing for cell size. If you have a larger cell, the daughter cells are also going to be larger. The progenyles are also going to be larger, and what they're kind of saying here is that this is not a random process. This is very specific and it may have a specific functional reason, but we don't want to go there necessarily um so upon nine and down regulation.

C

This variability propagation is lost, an outlier frequency and variability in cell cycle durations in lineages reduced. So.

B

There's the cell.

C

Cycle duration aspect, so the cell cycle duration you'll, see this in the C elegans lineage tree sort of later on, at least by the 256 cell stage. You start to see these asymmetric lengths of division. So up until that point, the divisions are relatively constant, but they start to get. They start to see variability, and this is part of this priming that we talked about in the last paper. What this is also due to some of these um asymmetric divisions.

C

So this is actually talking about down regulation. As you know, as in the progeny as the lineage continues, this variability propagation is eventually lost in outlier frequencing variability is reduced, as external influences are not detectable. We propose that a cell autonomous process, possibly involved in cell specialization, determine cell cycle duration variability.

C

So this is where we're talking about um sort of the the fate of the cells it's being primed as it becomes a terminally differentiated cell, uh there's a cell autonomous process that takes place, and so that's what they're talking about this paper- um and so uh you know this. This kind of goes through what we've been talking about, um and so, let's see.

C

So one of the questions they ask is we wondered whether the variability in cell cycle duration of related cells originates from intrinsic processes, segregating asymmetrically in divisions, our working hypothesis, so they're working hypothesis was that the variability in cell cycle duration of related cells, so there's going to be this asymmetric division. There's going to be this variability in cell cycle duration, does this originate from intrinsic processes, segregating asymmetrically in divisions and that's what they're they're kind of uh that's? What they're kind of expecting test this? We searched for reoccurring cell cycle motifs in the lineages.

C

First, we computed all possible mother-daughter pairs, sibling cell pairs, but your daughter and ordered pairs who pairs ourselves among granddaughter cells.

C

So if you look at the this in terms of a lineage tree, yeah three sets of relationships, we have a mother-daughter relationship, which is where we have a cell and we have two. We have a mother cell. We have two daughter cells, then we have four granddaughter cells.

C

That are like this sort of you know this third generation, and then you have pairings. So we have the mother-daughter pairs, which are two, so you have a mother and a daughter here and a mother and a daughter here you have the daughter, daughter pairs, which are these two daughters and then the granddaughter appears and so I guess you could take these like combinatorially if you wish, but that's what they look like.

C

um We investigated with our cell cycle durations between the members of each of these pairs, or rather the same or very different from each other.

C

This was also done in Sandler at all reference 21, the similarity, the cell cycle durations is highest between daughter cell pairs, followed by granddaughter cell pairs and the least between mother and daughter cells, as reported.

C

So this is like where you have these uh this variation. So it's where this effect wears off over time. That's what they're, finding um and said the similarity of cell cycle duration sites when you build our cell appear as well by granddaughter cell pairs. Actually, that's not what they're saying what they're saying is that these similar there's a similarity between the granddaughter cells, similarity between the daughter cell Pairs and then the least between mother-daughter cells, is because it's an asymmetric division as cells of the same generation range are those more likely to be similar.

C

So if you're in the same generation, you should be more similar diversity. Motifs among these cells are particularly meaningful further, as diversity is among Brandon or cell pairs, as this is higher than among daughter cells. We looked at whether these pairs form special motifs.

C

We therefore generated all granddaughter sets a set of four cell cycle, durations of the granddaughter cells of the same grandmother's cell and analyze, whether these four cells are always very alike or show relative to each other, reproducible, diversity, motifs in their cell cycle durations, but theoretically, most minimalist diversity, motifs among the cell cycle. Durations of four cells are the following two, so what they've done is they've taken these groups they've looked at their sort of similarity if they're similar within groups between groups, they're least similar.

C

So if, as a little bit of a confusing point because you're doing this analysis of comparisons, but basically what they're arguing is that in each of these groups, you have motifs, you have these diversity motifs, so the first one is a three to one Motif or one of the cells has a very different cell cycle duration. That's the outlier cell relative to the other three.

C

So if we go back to the granddaughter cells, you see that they're four cells and those four cells there's a similarity that marks, maybe three of them and the fourth one actually shows an outlier Behavior. So it might take longer to divide. So that's where you're getting this asymmetry.

C

So that's one Motif, then the other Motif is a tube to two Motif, whether where the cell cycle durations are more similar within a pair than between the pairs. So this goes back to the granddaughter cells. Again you have. These are the pairs, the daughter, the granddaughters from a specific daughter, and so those two would be comparable, and so the two to two means that uh two of these will be outliers compared to the other two. So that means that the sort of what the daughter, one of the daughters, is passing down this timing difference.

C

The timing. Difference will be the same in both granddaughters, say from daughter, one and daughter two it'll be the same, but between the descendants of daughter, one and daughter two they'll be different, so that's what they mean by diversity, motifs and by outliers, the behaviors and outlier. If it's different from the norm- and you know it happens in these different ways- recall that all the granddaughters are descended from the same Mother cell. So they all inherit uh similar things.

C

It's just that this change happens as sort of an outlier, so the asymmetry is an outlier Behavior to capture our lineage is such motifs that are based on relative differences. Within a granddaughter said. We use the outlier detection method in this case the four cells- the granddaughter set, cannot be separated into two clusters. With a given threshold, Memphis sends the granddaughter set of flat motif so they're interested in this three to one diversity motif um and Bismol.

C

Chief is also remarkable as one of the two daughter cells that gives rise to the 3-1 granddaughter set divided into an outlier and a non-outly, or so the throne Motif occurs in two manifestations.

C

The first is the three L motif where one of the cell cells as a long cell cycle duration, so a long cell cycle duration is basically where the division time is longer. So cell cycle is really setting a lot of the gain for this process. So let's take an example: here we have two daughter cells called her one and daughter two and from daughter one. We have two granddaughter cells.

C

A

Then those cells.

C

C

So daughter, one divides at a certain rate, and this is maybe different from daughter, two or maybe the same from the mother and then the sales and granddaughter two they're they're progeny you're, dividing into surgery.

B

We're not real.

C

Interested in that right now we're interested in granddaughter one and granddaughter too, and, as you can see here, there's a difference in time, a length of time which these descendants are generated from each granddaughter now in granddaughter one and granddaughter two. This time is dependent on Cell cycling, so there's a whole circuit, there's a whole grn as it were for a cell cycle and regulating cell cycle, and so that period is going to increase in granddaughter one versus granddaughter two: that's what they're talking about with this increase in cell cycle?

C

It's just the amount of time it takes for the cell I could go through a division, and so obviously, if it goes through a shorter division, it can produce progeny more quickly.

C

So this is where one of the cells is a long cell cycle. Duration relative to the other three, and so in this case, your three to one Motif is that all of the cells, even the other granddaughters of D2 here, have the same length of division, short divisions and then gd1 over here.

C

So this is three and four gd1 over here is a long division. That's what they're talking about that motif and then through the 3s Motif, is where one of the cells uh has a shorter cell cycle. Duration, s cells relative to the other three, and so this is where it's the opposite of this, where SD or gd1 would have a short cycle and the rest of them have a long cycle.

C

So this is, you know this is all dependent on cell cycle and modifying that circuit, modifying the grn, and so this is. This is part of this whole model, specifically each granddaughter said in the 36 SI non-target lineages was ordered by magnitude of the cell cycle duration and then hierarchical clustered based on and how uh mahalan avoid distances. So this is a method where the you know evaluate the clustering method, so.

A

C

Is actually I think where they're doing this clustering um and they show these different cells?

C

um So if I'm not mistaken, this is actually tracked image lifetimes of cells, with indicated cell identities originating from a single cell over 75 hours, represented in a lineage tree nodes, depict cell divisions, Edge length reports tracked image lifetimes bijectively assigned sub-identities corresponds to cell cycle durations of these cells if the cell could be imaged over the entire lifetime. So basically, this is a the length of time for the cell cycle.

C

These are the different cells and you can see this asymmetry or margin where time you have a symmetry here and then you have Slade asymmetry here and a longer period of asymmetry here, and so you can see this compounds across these different sublineages where you get these asymmetries introduced, and so this shows that this is a compounding process, and this is kind of a way they have to do this from Imaging, um but they were able to reconstruct this.

C

So this shows the diversity motifs this figure. This is actually where they show some things. I was diagramming. This is the long division. This is the short division and you notice it's one of four granddaughter cells. uh This is the two to two where you can have it in you know from one daughter or even have an asymmetrical division in both daughters. So in each case here a daughter one produces a this.

B

A

C

Diagram originally, but it didn't I, didn't understand what they were getting at here: um E1 and D2. So this is like a short a long from each daughter, and so this is again another uh another way that this can occur. Then, there's of course the flat. Where there's no variation, then they should some of these experiments where they actually get an empirical measurement in the experimental data, and then they permutate these trees to show some of these motifs, and so this is where they're getting some graphs from some of these experiments.

C

um Then they show this dissimilarity index versus frequency, and so you can see the empirical examples here. This is a nine in which they referred to in the abstract.

C

An easy difference is you can see that there's a lower frequency in the nine in uh and higher frequency in the non-target, and then of you know you have this flat expression here, where the in the flat case, where there's no asymmetry and so there's it's quite a bit different.

C

So this is, you know they kind of walked through this. This is a nice simulation of asymmetries, and things like that- um and this is uh so now I talk about the non-stochastic distribution of S and L cell divisions.

C

So we asked whether a cell intrinsic process or noise contributes to the generation of the three to one Motif. So is it just the noises like causing one of these to be short, one of these uh cell division lengths to be shorter than the other three, because it seems like that might be a possible thing that you know noise could affect it so that you maybe have one outlier out of four or one outlier out of 64..

C

um So this is an important question to ask. At this end, we probe whether a cell with an outlying cell cycle duration propagate some effect onto the cell cycle, durations of its two daughter cells in the Next Generation, thus affecting the three one Motif frequency in the Next Generation.

C

So to do this, they analyze the positions of the relative outlier cells and the lineages, comparing them with all the other cells terms. Zero cells and following propagation tests reveals whether cell cycle, duration of daughter cells or uniformly at randomly linked to Mother cells, which we could Define here as a stochastic Association or a non-uniformity axis, which we Define as a non-stochastic association between mother and two daughter cells in terms of cell cycle duration.

C

So they basically found that there's this absence of detective detectable external influences. This is uh uh not an extrinsic thing, it's an internal thing or an intrinsic thing, and this this expression of this nine in a racist statistical influence of parent cells. So this is something that I've also um show in this figure.

C

So, let's all for that paper, uh uh thank you for paying attention. Hope you learned something.