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.
0) What I already know:
ReplyDeleteThe difficulties involved with dealing with nonlinear, stochastic biological systems, most of the dimension reduction/machine learning techniques referenced, reductionism vs. systems biology, the microfluidic control system and the problem being posed with using metabolomics to observe cells.
1) The most important thing I learned:
I have further appreciation for the complexities of human biology and a better comprehension of the tools and approaches presented now that I have seen these techniques in different contexts.
2) My most pressing question:
For the PDEs that appear in biological models, I would expect most of them to be due to spatial variance. Are there any common causes PDEs to form in a model besides spatial variance?
How do you read a S-Plot and what kind of information does it give?
For the system using symbolic regression, what kind of building blocks are commonly used for biological systems? How is overfitting avoided? How is model complexity limited?
I get the rough idea of the P vs. NP problem but I do not have a full appreciation for it. Why is this problem so important?
3) Suggestion for class presentation:
P vs. NP problem
Symbolic Regression
Dimension Reduction Techniques (PCA, Kernel PCA, t-SNE, SOM, ...)
Machine Learning Algorithms (Supervised vs. Unsupervised, Clustering, Classification, Modeling,...)
Partial Differential Equations and how they can be approximated
4) Thoughts on class or PPT:
This PPT shows the complexity of human biology and a method for approach such complex problems. It serves as an example of how different philosophies/ideas and tools can be amalgamated to approach complex problems from a new perspective.
0-1) Good
Delete2) I know only of PDEs that describe spatial information
S-plots are not as intuitive as volcano plots, which are now more common. Check out https://en.wikipedia.org/wiki/Volcano_plot_(statistics)
If you read my paper with Schmidt as the first author, you will see examples of effect of different starting seed models. There is a later paper by Nemenman that cites it and uses different mathematical forms to fit the data. Basically, the closer you start, the better you do.
The categorization of problems as NP, etc. is a general measure of the complexity of the problem and the resources needed to solve it as it scales from small to large.
3) Good suggestions - it's not clear whether we will have time for these this semester.
4) Good.
3)
0) What I already knew
ReplyDeleteIn vitro experiments are on cells that often have cancer and are much simpler than biological organisms. They are not representative of complex organisms.
1) What I learned
The cell media problem is interesting. I am currently running out of media for my cells because I use about 20 mL each time I passage them and I passage them every 3 days. I enjoyed reading about how the microformulator enables faster mixing for experiments in 96 well plates and reduces amount of cell media used.
2) Most pressing questions
Slide 42: Not all organs in the human body can be included in the organ on a chip, so which key organs should be included? How are missing secretory organs substituted by the missing organ microformulator? I would love more clarification.
3) Suggestion for clas presentation:
Which key organs should be included in the organ on a chip?
4) Thoughts
This presentation shows some of the shortcomings of current in vitro research and how microfluidics can be used to solve these problems. This is particularly interesting to me because my research lab tests nanoparticles on cells and in mice to validate drug delivery mechanisms.
0) Good
Delete1) I'm glad you appreciate the problem! Can you mix up media by the liter? Or does it not keep? The uF is being designed to simplify the problem, particularly when lots of different formulations are needed.
2) Remind me to send you the Cyr article.
3) This is determined by the question you are asking. Check out ADMET
4) Parts of this should be quite relevant to what you are doing.
0.) What I already knew:
ReplyDeleteBiological systems contain an incredible amount of integrated parts and information. Understanding these systems presents many challenges. Knowing the values of parameters may take a large body of work. In addition, storing and interpreting all of this data offers yet another challenge.
1.) Most Important Thing I learned:
The future of systems biology may come in the form of topological analysis of data. The complexity of these systems may escape human understanding.
2.) Pressing Questions:
I am confused by slide 47. What does "functional annotation is not keeping pace with gene discovery" mean? Could sensitivities to initial conditions (classical chaos) be lost in determining these systems? Could ontological uncertainty play a role in these systems and does this invalidate ODE/PDE models?
3.) Suggestions for Class Presentation:
Topological analysis of data. Symbolic regression.
4.) Thoughts
The presentation is a lot of information to take in at once and I'm not sure if I fully grasped enough of it.
0) Exactly
Delete1) Yes!
2) People are identifying new genes much faster than they can determine what the gene is doing (i.e, function)
3) Yes, but it is not clear that we will have time this semester.
4) Read it a second time!
0) What I already knew:
ReplyDeleteI have had previous experience in Genetics studying some of the complex regulators of gene expression, as well as some experience with a lot of the cell biology discussed in this powerpoint. For example, I have some background in topics like G-protein coupled receptors, gene transcription, and intracellular biochemical signaling. However, it was really interesting to visualize these cell signaling networks as electrical systems with actuators, controls, and sensors serving the roles of many proteins and molecules.
1) Most important thing I learned
One of the most fascinating things I learned that I had never previously considered were the various time scales of systems biology. From the nanosecond with ion channel gating to 10^9 second for aging, it was really interesting to get a figure that put in perspective the relative time scales of various biological phenomena.
2) Most pressing question
How exactly are epigenetic networks modeled in systems biology?
3) Topic for class
Developing a toy model or molecular wiring diagram for the immune response and immune cell proliferation
4) Thoughts for class
I am really interested to see how systems biologists continue to tackle the now data driven domain of biology. I'm curious to see what new technology is developed to analyze the various "omics" of biology.
0) Excellent!
ReplyDelete1) Time scale and dynamics are incredibly important!!
2) We need to discuss this in class - on assignment list.
3) If you are interested, this could be a good exercise!
4) We will see more of this throughout the class.
0) What I already knew:
ReplyDeleteThat the well-mixed approach to modeling cells is not very accurate. That modeling of differentiation and epigenetics involves stochastic modeling and basins of attractors for studying gene regulatory dynamics. That due to the scales of complexity in cells more bottom-up approaches to modeling are needed and that microfluidic encapsulation of cells allows for near single cell measurements.
1) Most important thing I learned
Organ on a chip with Actuators, Controls, Sensors, and effluent analysis can allow for Automated Omni-Omics to collect data and generate models and narrow down the models that describe the system. That Multiscale interactions between the different length scales in biological systems produce emergent phenomena on top of interactions due to reducing larger scales as many smaller scale networks connected through adjacent. And that while genomic data is rapidly increasing the information on the biochemical reactions and interactions of these proteins are stagnating.
2) My most pressing questions
How do you utilize UPLC-nESI-IM-MS cool to dissect the behavior of cellular signaling networks and integrate that data to
characterizing cellular networks and models?
How does model parameter sensitivity change as the amount of parameters in your model get much larger?
Why are computational geometry and typologies with high-dimensional data approaches better than model simplification through dimensional reduction?
3) Suggestion for class Presentation
1. Modeling Chromatin remodeling effect on gene expression and Systems Biology of cell identity
2. How are transient behavior influence feedback and affect cellular dynamics at some scale in biological systems
4) Thoughts
Fisher's questionnaire approach seems interesting to me but I am not sure why it is better than non-randomized design of perturbations that are picked based on how likely they are to maximize the information obtained(Bayesian approaches). Philosophically I not sure it is useful to think that systems are too complicated for humans to fully comprehend and think it is better to think of these as thresholds that need to be meet before such an understanding can be made. I Really like the robot scientist approach. It really seems close to some research ideas I have on going towards automating the scientific process. Where Computational methodologies and robotic technologies can be leveraged to automated model design, validation on top of automating generation of hypothesizes, experimental design, execution, and data analysis using approaches in machine learning and AI development.
Kylie Balotin
ReplyDeleteReading Assignment 3: Intro to Systems Biology
0. What I already knew
I knew most of the biology described in the powerpoint. I’ve seen the slides from Huang on HL60 differentiation to neutrophils. I know about microfluidic devices and the microformulator.
1. What I learned
Other than the projects that I am/have been involved with, I didn’t know too much about VIIBRE, so it was interesting to learn more about the projects that your group does. I also learned what ontological failure and epistemological failure. I knew about these types of failure, but I’ve never heard this terminology before.
2. Most pressing questions
There are so many possible output sensors, and it seems like it could get pretty expensive and time consuming to cover all the possibilities. How many sensors would be “enough” to describe a system?
3. Suggestion for class presentation
It took me a few classes to understand the Waddington’s epigenetic landscape, so it might be useful for some of newer members of the class to get a presentation on this concept. However, I am not sure if this class will be as focused on stem cell differentiation as the previous class was.
4. Thoughts on class or PPT
I think the figure on slide 21 really shows the complexity and how the interactions between the scales can be an issue when trying to study and describe a system.
0) Knew: The epigenetic landscape of cells consists of "downhill potential" for cells to choose their fate, but the landscape is constantly changed and regulated by the cell's environment.
ReplyDelete1) Learned: Models of biological networks may eventually get so large that we either can't build them, or we can no longer interpret what they tell us.
2) Questions: Slides 64 - 69, what does it mean that MEDI metabolomic data is "self-organizing" and what does it tell you about the samples?
3) Presentation: I like the discussion of instrumentation methods that allow for multi-omics measurement of in vitro systems. I think it would be interesting to go more in-depth about different methods, their strengths and weaknesses, and how they contribute to the available data that drives model-building and interpretation.
4) Thoughts: I like the intersection of "philosophy" - how to think about tackling systems bio problems - and "strategy" - how to actually collect data and build models that are reliable and accurate representations of the system in question.
0) What I already knew:
ReplyDeleteThis always feel like a hard question. At first I seem to know a little something, but I soon realize I don't REALLY understand anything.
1) Most important thing I learned
Modeling a single mammalian cell requires more than 100,000 dynamic variables!? Insane.
2) Most pressing question
Has there been attempts to ambitiously model a mammalian cell with more than 100,000 dynamic variables??
3) Topic for class
Modeling epithelial cell layers and what abstractions can be made.
4) Thoughts
Emerson M. Pugh's quote reminded me of a quote by Piet Hein: "The universe may be as great as they say. But it wouldn't be missed if it didn't exist." There's also something more original that I say sometimes, "trying to understand how my body works is like trying to pick myself up".
0 Knew: I know biology has nearly bottomless complexity, and the "rules" as we continue to understand them often have exceptions. Given this complexity, it is an attractive notion to reduce the variables within a given system in order to more easily tease apart the nature of each individual interaction. The trade-off to this reduction of variables is often the removal of factors that may be crucial to a biological outcome, like the microenvironment. Biomimetics and modern experimental/computational/modeling approaches are now allowing for more of these features to be built back into experimental approaches while giving researchers the power to measure and qualify the individual contributions and outcomes of an increasing number of variables.
ReplyDelete1 Learned:
ODE vs PDE distinction: This was mentioned in a slide and upon looking it up and finding it mentioned in the nonlinear ODEs paper I think I'm getting the idea of it.
It seems like PDEs can overcome some of the shortfalls of modeling with ODEs, at the cost of added complexity. I can't easily look at the math and understand it but in principal I like the idea; it seems like a means of giving a variable dimensional wiggle room in which variations can be accounted for and perhaps recognized across multiple variables - or perhaps a way of accounting for unknown variables in general. Whatever the case, building a vocabulary and rudimentary understanding of the diversity of approaches is helpful.
2 Pressing ??: To what extent can models be stacked or embedded within one another? (i.e. ode networks modeling individual cells interacting in a network that ultimately models their expression patterns at the tissue level)
3 Presentation: gut biomimetics/culture/coculture systems
4 Thoughts:
This presentation sets up the problem of complexity and offers some tools to unravel it; for this project one important takeaway will be maintaining a balance between over-reduction and over-complication. How do we design an experimental system that captures the complexity of the gut epithelium, its associated stroma, microbiota, and other luminal contents without over-convoluting any measurable variables? It's perhaps analogous to studying a rainforest by selecting a handful of plants and animals and putting them in a terrarium. Unless we pick plants and animals that can compliment one another and fall into a balance of sorts, we haven't come close to approximating an ecosystem, just a zoo exhibit. Did you know that in an environment with no wind a tree's limbs grow brittle? What variables may be important within our system?
0) KNEW
ReplyDeleteI knew that biology is complex, even from the scale of systems biology. It is the interactions (spatially and temporally) that cause a lot of this complexity. I also knew a bit about automated biology (from the class last year, of course), including robot scientists and the microformulator. I knew that most biological phenomena are stochastic, dynamic, and/or non-linear. I also already knew about epigentic landscape. I was aware of the differentiation of neutrophils, from last year.
1) LEARNED
The yeast cell cycle is the most well modeled signaling system. BioMEMS (microfluidics) is amazing.
2) QUESTION
From slide 28 "How do genome-wide expression profiles change
coherently from one stable pattern to another?"
3) PRESENTATION
HL60 to Neutrophil paper?
4) THOUGHTS
I liked this presentation. Very thorough for sure. Lots of figures that I felt like were a bit too complex to all understand in a short amount of time, but I definitely got the big picture.