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.
Wednesday, October 9, 2019
SysBio19 Asgn_1_Class__Joo_article_2019_10_10
Read carefully Joo article Probabilitic Boolean Networks. Answer the
questions: 1) What did you already know? 2) What did you learn? 3) What
is your most pressing question that you would like Kyle to explain in
class? 4) How does this relate to your project?
Subscribe to:
Post Comments (Atom)
1) What did you already know?
ReplyDeleteI have read this article before so I knew most of the things with the article but it has been a while so I am fuzzy on the details.
2) What did you learn?
I relearned all the metrics for relative dynamic stability, especially the one proposed in the paper, the stability index.
3) What is your most pressing question that you would like Kyle to explain in class?
N/A
4) How does this relate to your project?
This is my research.
1) What did you already know?
ReplyDeletethe low efficiency of cell state transitions is poorly predicted by linear pathway diagrams involving "fate-determining" master regulators
2) What did you learn?
general concept of using Boolean networks to model GRNs, with attractors defining cell phenotypes, and difficulties in accurately representing quasi-potential in these networks
concept of metastable intermediate states- relevant when talking about epigentic memory during early passages of "iPSCs" that may actually be more parent cell/stem cell hybrids. although these are represented as attractors however they are described as "metastable" so i dont fully understand this
3) What is your most pressing question that you would like Kyle to explain in class?
-dont fully understand all of the methods of calculating relative dynamic stability of attractors other than basin size(SSP, MFPT, BTR, and SA (and how 1-degree neighbor distribution plays a role))
-how do you transition from one attractor to another? my basic understanding of an attractor state is that it is a steady state/cycle so not sure how you escape this, or is this where the quasipotential/probabilities come in
4) How does this relate to your project?
could try to create a network involving genes that regulate dedifferentiation from a certain parent cell
-dont fully understand all of the methods of calculating relative dynamic stability of attractors other than basin size(SSP, MFPT, BTR, and SA (and how 1-degree neighbor distribution plays a role))
DeleteThe SSP describes how likely it is for the network to change state after it has reached steady state. The approach uses Markov Chains and hamming distances to calculate the probability.
MFPT is the average time it takes to go from one state to a different state through a stochastic process. This uses transition probabilities, Markov chains and the SSP for the calculation.
BTR uses Monte Carlo simulation with the state transition matrix to calcuate the likelihood of transitioning to another basin of attraction.
SA aims to count the number of Hamming neighbors that stay or leave a basin.
Refer to the methods section for more details.
-how do you transition from one attractor to another? my basic understanding of an attractor state is that it is a steady state/cycle so not sure how you escape this, or is this where the quasipotential/probabilities come in
Stochastic noise drives the transition from one attractor to another.
1) What did you already know?
ReplyDeleteyou can model relationships between genes through Boolean networks
2) What did you learn?
Mathematical methods for assessing stability of states in a graph
3) What is your most pressing question that you would like Kyle to explain in class?
How can this Boolean network change if you use a few modules of genes rather than the four key genes that they used?
Rather than looking at the discrete states that the researchers manipulated in order to analyze stability, how can the entire quasi-potential landscape be modeled so that one can simply find the minima of the landscape?
4) How does this relate to your project?
For the stromal to iPSCs project, we could do a literature review to generate a table of known interactions between differentiation factor genes and use that information to generate a boolean network.
3) continued
DeleteHow do you determine what the quasi-potential of a state is?
How can this Boolean network change if you use a few modules of genes rather than the four key genes that they used?
DeleteAdding more genes would decrease your signal-to-noise ratio since you are potentially adding in irrelevant genes to the process. Further with larger networks, it is more difficult to analyze or fit Boolean Functions since the state spaces grows as 2^n where n is the number of vertices in the network.
Rather than looking at the discrete states that the researchers manipulated in order to analyze stability, how can the entire quasi-potential landscape be modeled so that one can simply find the minima of the landscape?
You can use an ODE model to try to elucidate this but this should be taken with a grain of salt since you will have to decrease the size of your model so you can fit rate constants.
How do you determine what the quasi-potential of a state is?
You assume the driving force, as modeled by your ODE system is conservative which means F=-dU/dx and integrate your equations to get the quasi-potential. It is called a quasi-potential because the system is not truly conservative meaning there are path-dependent forces in the system. Refer to https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3481575/ for a more thorough explanation.
1) Not much. We've studied network approaches to describing GRNS, but not in terms of Boolean networks.
ReplyDelete2)I learned that Boolean networks were an option for describing Gene regulatory networks and its utility in determining attractor states (which in their case we're different that expected from experiment).
3) how did they implement the network? I think I understand how they generated the specific network for this system, but I'm still fuzzy on its implementation. I don't fully appreciate how their choosing interaction rates and how they connect the various cell states (ie mesenchymal, hybrid, and epithelial) together.
4) The cardiac project looks at the transition from mesenchymal cells to either fibroblasts or cardiomyocytes. We could create a Boolean network for that system regulated by TGFB similar to that generated here.
how did they implement the network?
DeleteThe inferred functions from data and then used a Markov Chain to determine the basins of attraction.
I think I understand how they generated the specific network for this system, but I'm still fuzzy on its implementation. I don't fully appreciate how their choosing interaction rates and how they connect the various cell states (ie mesenchymal, hybrid, and epithelial) together.
The steady states can be inferred from what is being up or down regulated in the steady state. Their data will tell them how to label the attractors.
1.) What did you already know?
ReplyDeleteI was generally aware of the EMT transition under the context of metastasis. With many cancer cell types undergoing at least a partial EMT phenotype change prior to migrating away from the primary tumor site.
2.) What did you learn?
Transition probabilities can be determined using these boolean networks. And from the transition probability, we can determine which cell attractor is now the most likely.
3.) What is your most pressing question that you would like Kyle to explain in class? What do these boolean network models look like in practice? How do you interface with the model?
4.) How does this relate to your project?
While this specific network pertains to EMT, the general model can be applied to other studies of cell transition. Thus we may be able to take advantage of this system in cardiac reprogramming. Also there may be intermediate cell transition states on the journey from cardiac to ipsc or vice versa.
What do these boolean network models look like in practice?
DeleteIn practice, a boolean network is a set of vertices (genes) with defined edges (interactions) connecting them. The model has a set of boolean functions to define the interactions and is given a time component through a Markov Chain.
How do you interface with the model?
You change the model through defining your network and boolean functions. You can alter the trajectory by allowing noise or perturbations to induce transitions to different states.
1) What did you already know?
ReplyDeleteCell states are generally unstable, and their stability is dependent on the current state.
2) What did you learn?
Most of the math is new to me. I thought the Hamming distance is a useful tool and way to visualize how the network is determined.
3) What is your most pressing question that you would like Kyle to explain in class?
How is basin size determined?
4) How does this relate to your project?
induced pluripotency often involves an EMT transition, so this model can help predict better targets such that we get higher efficacy induction.
How is basin size determined?
DeleteBasin size is determined by the size of the basin of attraction, which is the number of states in a given basin of attraction.
1) What did you already know?
ReplyDeleteI understood what they are talking about with regards to dynamic nature of cells states, the EMT progression
2) What did you learn?
The utilization of boolean models to understand state stability was something new to me.
3) What is your most pressing question that you would like Kyle to explain in class?
- How does the quasi potential landscape show the different attractors? What exactly does it show/mean? Is it the same idea as that waddington surface?
-Why are boolean networks good? What do they do that other models don't?
-How do you generate the basins and figure out the sizes?
-What is the hamming distance?
4) How does this relate to your project?
We can generate Boolean models to understand cell differentiation. In the case of cardiomyocyte dedifferentiation, we can model the situation to determine an experimental procedure to manipulate state transition which could prove valuable in understanding and curing the problem.
- How does the quasi potential landscape show the different attractors?
DeleteEach potential well in the landscape is an attractor. The example has three potential wells so there are three attractors.
What exactly does it show/mean?
The quasi-potential landscape describes how a cell's type will change overtime. More pluripotent states are higher energy and more differentiate states are lower energy.
Is it the same idea as that waddington surface?
Effectively yes.
-Why are boolean networks good? What do they do that other models don't?
Boolean networks require less data, are less computationally intense and highly interpretable. They lack the predictive powers of other models but can help you understand a network's dynamics.
-How do you generate the basins and figure out the sizes?
The basins are made through a Markov chain and the sizes are the number of states in each basin.
-What is the hamming distance?
It is the number of components in two vectors that are different.
1) What did you already know?
ReplyDeleteThat boolean networks were a type of graph showing the states possible and transitions between the states and that cells when pushed out of their well will fluctuate in unstable states till reaching a stable state.
2) What did you learn?
I learned that a method I had learned primarily for designing circuits could be applied to biological contexts.
3) What is your most pressing question that you would like Kyle to explain in class?
Is this method possible for larger amount of variables or would it require too much data?
4) How does this relate to your project?
This allows for a new view at modeling the problem in an easier way that can be easily visualized when compared to just and epigenetic landscape. This will also allow for better calculations of well size and stability for the different cell types.
Is this method possible for larger amount of variables or would it require too much data?
DeleteYou potentially run into issues with signal to noise but I cannot find an upper bound on network size. You will need to have reliable data for your network to determine functions and if you are trying to determine network structure, that will require a ton of data.
1) What did you already know?
ReplyDelete* I knew that gene regulatory networks could be modeled as a set of Boolean equations for simplification and understanding.
2) What did you learn?
* I thought it was interesting that they approached cell type from the definition of the most stable combination of gene networks combining to result in a stable state.
3) What is your most pressing question that you would like Kyle to explain in class?
* How does this explanation of cell phenotype relate to cancer? Are cancer cells more stable?
4) How does this relate to your project?
* From a perspective of cell differentiation, it's useful to consider the cell phenotype as a stable alignment of genes not only because it gives insight into common motifs in gene expression, it can also help us understand specifically what genes drive alignment. By this I mean, which genes or transcription factors are either most selective in their alignment or drive alignment trends? I would like to target those pivotal genes for further analysis.
How does this explanation of cell phenotype relate to cancer? Are cancer cells more stable?
DeleteEMT is an important process for cancer. Cancer can be thought as being in a malignant attractor state which is highly stable.
1) What did you already know?: How boolean networks are generated, truth table and definition of states.
ReplyDelete2) What did you learn?: They picked their boolean network by generating all possible boolean functions with 5 gene nodes, then fit to observed cell states and attractors.
3) What is your most pressing question?: How well do these methods scale with the size of the network? 5 genes seems oversimplified compared to the types of GRNs we are hoping to work with.
4) How does this relate to your project?: Boolean or agent-based deterministic models could be useful for a well-characterized developmental system such as the intestinal crypt.
How well do these methods scale with the size of the network?
DeleteThe state space scales like 2^n but I cannot find the bound on Markov chains. Also with larger networks you have less signal-to-noise.
5 genes seems oversimplified compared to the types of GRNs we are hoping to work with.
The aim with simplifying the model is you see the core behavior of the network dynamics. It won't be 100% accurate but simple models are robust.
1) boolean basics, gene networks can be modeled with boolean logic
ReplyDelete2) How boolean networks can be used to define stable states and define how much perturbation needs to be done to change wells and how likely is it to find a cell in a particular state
3) How is data incorporated into a boolean network. Do you look at transcript data and form connections?
4) a boolean logic map for each disease state might give us more information into drug design to move from a disease state to a healed state in a cell or tissue
3) How is data incorporated into a boolean network. Do you look at transcript data and form connections?
ReplyDeleteData is used for determining network structure and boolean functions. Transcript data would be used to determine gene expression state.