The Systems Biology of COVID-19 and the SARS-CoV-2 virus. The class will build a foundation that includes the emergence of complexity, simple biological subsystems, their reductionist and equivalent toy and organ-chip models, and the measurements required to specify model architecture and parameters. Applications to biology, physiology, medicine, chemical and biological defense, pharmacology, drug discovery, and toxicology. UGrad: PHYS 240 01 and BME 290B; Grad: PHYS 326 and BME 395C.
Monday, September 26, 2016
SysBio16 Asgn_11A_Class_11_Article_11_2016_09_29
Epigenetics/Attractors: Read Article 11 S. Huang, G. Eichler, Y.
Bar-Yam, and D. E. Ingber. Cell fates as high-dimensional attractor
states of a complex gene regulatory network. Phys.Rev.Lett. 94
(12):128701, 2005.. This is an introduction to both cell reprogramming
and data visualization. Post a PCRC.
Subscribe to:
Post Comments (Atom)
0. Knew:
ReplyDeleteI knew somewhat about attractors from previous discussions in class.
1. Learned:
The theory proposed at the beginning of the paper is that attractor states of the phenotypic state over time (S(t)) represent various differentiated cell types. This theory has not been experimentally tested yet. The goal of the experiment presented in the paper is to “obtain evidence that a stationary phenotype state is a high-dimensional attractor state”. This was done by using human promyelocytic HL60 cells that can be triggered to assume a stable state by a variety of means. This is important because one characteristic of an attractor state is that the state must be obtainable through many different paths.
Two paths were considered in the high-dimensional space; one triggered by atRNA and the other triggered by DMSO. The two states were monitored with a time resolution of 2 hours and then daily after the first day. The results are that even though the two trajectories exhibit very different gene expression patterns at first, the patterns converge to virtually the same pattern around day 6.
The main take away for me is that we can determine properties of complex networks, such as attractor states, without knowing the specifics of the network itself.
2. Pressing Questions:
I need more understanding behind the value b(t) and how it relates to bstat. I know that the values of these are what drove the motivation to filter out some of the genes considered in the experiment, but I do not think I am fully grasping this.
Another question I have is that is it really enough to consider only these two trajectories? Should more trajectories be considered before we can truly label a state as an attractor state? Are there other ways to test if a state is an attractor state?
3. Presentation Topic:
If there are any other examples of this method being used, that would be a nice to include in a presentation.
4. Thoughts:
This paper was easy to grasp, but I think I will learn even more about it from the class discussion.
I agree that more trajectories to the same point should be considered for the state space model to be convincing. I wonder if it's just hard to find different ways to induce the transition they were studying (neutrophils to HL60 cells). Also, is there some statistical analysis necessary to be certain that there is an attractor somewhere?
DeleteKelly McGee
ReplyDeleteArticle 11a
Cell Fates as High-Dimensional Attractor States of a Complex Gene Regulatory Network
0: Knew: I knew the basics of principal component analysis from having to learn for previous papers, additionally, I was at least somewhat aware of the complex issue of how cells differentiate into the many different phenotypes throughout the human body.
1: Learned: I learned a new use for PCA: studying how different differentiation affectors compare to each other across gene expression over time. I learned that there may be biological differences based in gene expression levels between cells that we have labeled as the same phenotype.
2: Questions: How were the principal components created and calculated? I didn't see an explanation anywhere.
3: Presentation: How we previously have distinguished cell phenotypes, so we have an understanding of how this may change that process.
4: Comments: Short, interesting, relevant paper.
Chinowsky_TheorSysBio_PCRC_09129016
ReplyDelete0. Knew
I have knowledge of how it is thought that phenotype follows from genotype, although there is some debate as to if that is entirely accurate. No matter what, phenotype is inherently connected to genotype, and the wide variety of genetic variation can account for the phenotypic variation that is part of what makes signaling networks so hard to construct.
1. Learned
I learned that the phenotypic state of a cell can be expressed, in terms of time t, by the activity of individual genes in the genome, creating a state vector. This vector can be used to represent the dynamics of a complex network within a cell. This article also looks at how we can follow different state trajectories through to differentiated states, specifically with the addition of DMSO and atRA in HL60 cells. The state vector methods allows us to follow these two trajectories through to different regions of the state space.
2. Pressing
Can this state space be expanded to include epigenetics? Can it account for epigenetics in the way it is outlined in this paper?
3. Presentation
A discussion on how genetics has evolved beyond genotype=phenotype
4. Thoughts
A lovely, succinct expression of a mathematical expression of gene regulatory networks and their relation to a final phenotype. Interesting and relevant.
0. KNEW:
ReplyDeleteI was familiar with the concept of attractor space, but more from a mathematical perspective rather than that of a theoretical study. In math, I had learned an attractor space is essentially the center point that a set of values revolves around. In my head I picture this as a body in space being gravitationally pulled to a larger body (the attractor). The route at which the smaller body takes around the larger body can be linear, or it can be multi-dimensional and chaotic, depending on what was happening when the attraction began.
1. LEARNED:
Learning about cell fate is trying to understand what causes a particular cell to develop into a final cell type. There are four cellular processes: proliferation, specialization, interaction, and movement, and there are techniques being developed to study cells as they differentiate into their final state. This paper suggests that attractor states can represent these various cell fates over time.
2. PRESSING QUESTIONS:
How many trajectories would be needed to be able to reasonably identify something as an attractor space? Two seems far too few in my opinion. I am confused by the principle component aspect of Figure 1B., further clarification would be appreciated.
3. PRESENTATION:
Any new work that has been published by this group since this paper was published in 2005.
4. THOUGHTS:
Looking forward to the group discussion.
0. Knew:
ReplyDeleteThe previous 2 class discussions really helped me to follow the procedures discussed in this paper.
1. Learned:
Learned a new technique (similar to others) to investigate regulatory networks and what an attractor state is in this context. I was familiar with different cell fates, but hadn't seen them referred to by attractor states.
Visualizing the self organizing maps helped my understanding a lot. Figure 2 clarified a some of what we've been discussing in class.
2. Pressing ?s:
Slightly confused by the disparity term b(t). I thought b was the intertrajectory distance. I could use a little explanation on how the intertrajectory distance is determined and exactly what it is telling us.
3. Presentation:
GEDI (is this similar to the function that goes through KEGG)?
I was pretty intrigued by the HL60 cells and though a presentation on cell characteristics that make them useful for investigating gene expression and networks may be interesting.
4. Thoughts:
I like the conciseness of the paper. But as the paper mentions, I was concerned with the small number of time points and know there's a lot that probably went unaccounted for in the time between time points.
0 Knew
ReplyDeleteConcept of cell multi-stability and attractor states
1 Learned
How to test an attractor hypothesis. Use of GEDI and PCA to reduce data dimensionality in this context.
2 Pressing
More time points and trajectories are needed for detailed state space analysis (esp near trajectory), so how does one determine the statistical significance of an experiment? Use Z-factor as in HTS? Can you assume a multivariate normal distribution to create a gaussian model, then use subsampling/bootstrapping to reassess normal assumption and subset interactions? Or create a Bayesian model?
3 Presentation
GEDI
Statistics in GRN
4 Thoughts
It complemented Huang's earlier paper (8A) "Dimensions of systems biology" well. I'm interested in seeing the approach applied to cells with more than two attractor states.
Asgn 11A, Article 11: Cell Fates as High-Dimensional Attractor States of a Complex Gene Regulatory Network
ReplyDelete0. KNEW
I knew that cells undergo lots of mRNA concentration changes when undergoing shifts in cell fates, and I knew we could monitor these changes using DNA microarrays. I knew how correlation analysis can be used to different the extent and significance of differences between two sets of data.
1. LEARNED
I found it super interesting that there is a cell lines that can differentiate to a specific cell type and that we know which chemical we can use to do that. I learned how they combined all of the relevant genes into an anonymous, combined vector. It was interesting to see that so much of the gene network was involved in these changes, but both process led to a very similar overall conclusion.
2. MOST PRESSING QUESTIONS
(Pg 2) I don't really understand what PC1 and PC2 are and how they relate to the analysis. What do they encompass and why are they used?
(pg 3) How many of these distinct cell states/fates have we found? How did we find them and do they all have such widespread changes in mRNA concentration?
What does the cell do when it gets signals for different cell fates at the same time? Does the cell have a way to pick which one is more important? Can cells go to different cell fates at the same time?
What details of the data is lost when we combine all of the mRNA states of an entire cell population into one huge equation?
3. PRESENTATION TOPIC
Further clarification on self-organizing maps, specifically on the SOMs in figure 1.
How do we reduce phase space (specifically by using principal components)?
4. THOUGHTS
I found the experimental procedure of this paper super interesting and ingenious. The idea of using two different chemicals to reach the same cell fate and then mapping the two paths to get to the final destination seems so complicated to come up with, but has such clean and understandable results. I'm curious what other results they would find if they did a pair-wise analysis for individual genes that canonically are important in the differentiation.
I agree that principal component analysis is probably an important thing for us to understand. It would be good for someone to give a brief presentation on it, and explain why it is helpful (for example, how it helps us visualize data from high-dimensional spaces).
DeleteBen Terrones
ReplyDeleteSysBio16 Asgn_11A_Class_11_Article_11_2016_09_29
0. KNEW:
Most of this was knew information although I have heard about the idea of cell fates acting as attractor states, just not in detail.
1. LEARNED:
I learned about the idea of cell fates as high dimensional attractor states. Researches tested the attractor hypothesis by using two distinct stimuli to trigger HL60 cells to differentiate into neutrophils and used microarrays to monitor gene expression patterns. They were able to show that the trajectories initially diverged but eventually converged to a common endpoint, a high dimensional attractor state. They concluded that a whole genome view is necessary because of the number of signals and perturbations that effect cells and the fact that they respond to this with only a few cell fates.
2. PRESSING ?:
Still kind of unclear about how the GEDI program works.
3. PRESENTATION:
GEDI
4. THOUGHTS:
Nice, short paper that got to the point.
Sylvia Morrow
ReplyDeleteAsgn11A_S. Huang, G. Eichler, Y. Bar-Yam, and D. E. Ingber. Cell fates as high-dimensional attractor states of a complex gene regulatory network.
0. KNEW: Slightly more than nothing about GEDI and self-orgainizing maps.
1. LEARNED: The basic approach of using a state vector to quantify, organize, and describe gene expression
2. PRESSING ?:
--"The number of time points monitored in the experiments also does not allow a detailed state space analysis as in other high-dimensional dynamic systems" (3) Is there a practical limitation that doesn't allow you to go back and get more time points?
--"We found more than 100 genes, including the neutrophil differentiation markers CD11b and G-CSFR1, that exhibited essentially equivalent dynamics in the atRA- and DMSO-triggered processes" (3) I don't understand this sentence well enough to understand why it excludes "an overall temporal shift in differentiation" (3).
3. PRESENTATION:
--Self-organizing maps
4. THOUGHTS: Nice paper. Compact, straightforward, and clear. All of the details that usually bog me down were in the supplement which I tried to read but didn't comprehend much of.
0. KNEW
ReplyDeleteI had a general idea about the attractor states of the cell and how forced differentiation in vitro is used to help identify gene states.
1. LEARNED
I have a better idea as to the change in gene expression due to differentiating stimuli over the course of days. I learned a little more about the terms used for GEDI, like partial convergence though it would be good to define it more rigidly. The researchers did not use any metadata in creating these genome-wide gene expression profiles. Instead of these discrete cell states differing by small regulatory circuits, over 90% of the genes are involved in these changes.
2. QUESTION
I am not super clear on why they used genes whose expression in atRA vs DMSO-induced neutrophils not considered significantly different were used for the analysis of trajectory (p.2)
3. PRESENTATION
GEDI
4. THOUGHTS
This was a short and sweet paper that was relatively easy to digest.
Class 11. Article 11 S. Huang, G. Eichler, Y. Bar-Yam, and D. E. Ingber. Cell fates as high-dimensional attractor states of a complex gene regulatory network. Phys.Rev.Lett. 94 (12):128701, 2005.
ReplyDelete0. KNEW:
I already read the longer paper first, so I understood the conceptualization of cell fates as balls rolling around in a high-dimensional space full of hills and valleys; accordingly, I understood that it is important to gather some experimental evidence to test the model. Also, I was familiar with principal component analysis from a paper we read last week, so I understood what was happening when it was used to visualize data (since 2773 dimensions are somewhat harder to visualize than 2 or 3).
1. LEARNED:
The idea here was to test the model Huang described in the other paper in a specific cell type transition: neutrophils (a kind of white blood cell) changing into HL60 cells (cells which differentiate into white blood cells). HL60 cells are analogous to stem cells here.
To test the model for state space, the transition was achieved in two ways: (1) using the chemical dimethylsulfoxide (DMSO) and (2) the hormone all-trans-retinoic acid (atRA). Expression levels for 2773 different genes were measured in the cells under study via DNA microarrays.
If the model is realistic, these diverging paths should converge on about the same point in state space.
2. QUESTIONS:
It seems like the experiment worked out okay, but it is clear that you need a lot more evidence than what you get can from studying one transition before the model becomes convincing. What experiments could be done to completely discard linear differentiate pathways as the explanation for cell fate behavior? Is there one? Aren't they, in some sense, two different ways of looking at the same thing?
3. PRESENTATION:
The other paper is a lot more important, I think, and should be presented first. Then it is clear why the experiment here is being done. After explaining the other paper, what's important is to explain the logic of the experiment here: they wanted to show that diverging trajectories can converge on the same point, which is characteristic of an attractor. This is less important, but it also might be nice for the presenter to explain why these particular cell types and this transition was chosen (was it just convenience, mostly?), and what might be difficult about the doing of an experiment like this.
4. THOUGHTS:
Supposing more evidence comes in (as it maybe has, since this was over 10 years ago), will experimental biologists actually embrace state space as a useful concept? Does it really benefit them to do so? What does it take to solidify a paradigm change like this if it's right?
0 Knew
ReplyDeleteFrom medical school I was familiar with APL and the role of ATRA in triggering cell differentiation to mature granulocytes. From the other Huang paper I was familiar with the concept of cellular state space. From my research I am familiar with GEDI and PCA, but I'm still learning the math behind them.
1 Learned
I learned that DMSO can trigger cell differentiation too. This is good to know, as DMSO is what we use as a vehicle (control) in my research. The concept of attractor states was also new to me.
2 Questions
In pg 3 paragraph 1 the authors write "PC1 appears to account for the expression change due to differentiation, while PC2 reflects the difference between the two trajectories" How did they figure that out? How do we go about interpreting 2 dimensional representations of high dimensional space? Is there a robust and consistent way of assigning points in 2D space?
What is represented by the two lines in figure 2a?
3. Presentation
Dimensionality reduction would make a good presentation. Perhaps limiting it to "how to interpret" representations of high dimensional space
4. Thoughts
Absolutely love this paper; best paper I've read this month.
I will be talking about PCA today so hopefully that will answer your questions. Maybe I didn't really understand this paper, but we reached very different conclusions; I thought the paper lacked substance, but this did happen in 2005.
Delete0. KNEW:
ReplyDeleteI knew what differentiation was and how cells all have the same genes. I understood the methods they performed including DNA microarrays. I knew what PCA was.
1. LEARNED:
I learned what is referred to as an attractor state. I learned that all the gene expression levels together can be thought of as a state of a cell and that this can change over time with phenotypic traits. I learned how exactly cells can be ‘programmed’ to differentiate. I learned that the method of inducing differentiation alters the cell differently, at least in the early stages.
2. PRESSING ?:
Has this gone further since 2005?
Page 94, p2: is 31% and 14% covering of variance ‘good’?
3. PRESENTATION:
I would focus on the self organizing maps and how they were created (i.e., GEDI).
4. THOUGHTS:
This paper was straightforward yet I don’t really get the significance. The idea of an attractor state, to me, just seems like a common state cells are in. Defining what that is instead of just confirming its existence would seem more exciting.
Class 11, Assignment 11a
ReplyDeleteCell Fates as High-Dimensional Attractor States of a Complex Gene Regulatory Network
(0) Knew:
I have knowledge of the process of differentiation in multicellular organisms (self-organization, self-renewal, etc). I also know that this process is governed by genome-wide regulatory networks that are manipulated differently in each cell type/layer.
(1) Learned:
I learned about the techniques the group used to evaluate gene expression at different timepoints using microarrays (in so doing I learned that DMSO/atRA can be used to differentiate HL60 progenitor cells into neutrophils). I also learned about the intertrajectory distance that the group used to correlate expression levels while excluding most of the noise produced by single gene measurements (~72% of the microarray set remain). I learned about the idea of cell fates as attractors (this is something I had previously given thought to).
(2) Pressing Questions:
I’d appreciate more clarification on the mechanism of a DNA microarray (they’re mentioned frequently in class, and while I understand the output they produce, I don’t quite understand how they’re prepared, and how they can measure so many signals in parallel and resolve them adequately). I’d also ask for clarification about how this method (and some of the mathematical techniques embraced by the group) can be used to develop a coherent network rather than isolated correlations – is this an Occam’s razor type situation? Can you infer relationships based on the least number of assumptions necessary to assemble a network from correlation information?
(3) Presentation Topic:
Time disparity of expression between microarray set and common set of genes? I’m a bit unclear on this
(4) Thoughts:
I’m impressed that this group devised a method to use genome-wide data to develop “globally coherent patterns of gene activation” and make inferences about the consequences of cell fate. Looking forward to seeing how this discussion proceeds in class.
This comment has been removed by the author.
ReplyDelete0.Knew: I knew that somehow, the thousands of genes in the genome can regulate each other’s activity. I knew that this regulatory network is present in cells of various different types and that yet, these cells are distinct because of the way the gene expression in each differs.
ReplyDelete1.Learned: I learned how to represent the phenotypic state of a cell as a vector of the activity of the genes of the cell, which makes sense because the central dogma says that genes direct proteins which direct gene expression. I learned that the theoretical result that gene networks have relatively few stable attractor states led to the belief that the attractor states are representative of differentiated cell types/cell fates. I learned about how studying the application of DMSO and atRA can help us understand genetic networks and pathways.
2.Pressing Questions: Where do HL60 cells come from? (p.1)
3. Presentation: a review of the experiment performed should briefly be presented
4. Thoughts: I think this article will make more sense to me with subsequent reads, but overall I found it very interesting. I appreciate mathematically representing the phenotypic state of a cell and the detailed explanation of the experiment performed.