The Biological Network

By Michael Begun

Networks have bloomed in the last few years, from “The Social Network” to the deluge of scientific papers studying such varied things as the financial system and genetic diseases through networks. While we have seen the term “network science” emerge, it is not exactly clear what this means. Wikipedia describes network science as an interdisciplinary academic field, but I think that networks are better viewed as tools for making sense of complex systems. Networks are one way that we can describe collections of interactions between people, websites, words, animals, genes, and so on.

For examples of popular network studies in biology, check out these papers: ecology, public health, and cancer genomics.

Research on social networks has a surprisingly long history in sociology. But the discussion of networks across diverse fields has taken off in the last decade, especially in systems biology. This is in large part due to computers, which allow us to construct and analyze huge networks. In biology, a major use of networks is to investigate gene regulation, uncovering relationships and hidden structure among genes. A thriving research area involves inferring networks from data such as E-MAPs.

The most exciting thing about biological networks, in my opinion, is that they allow us to examine large-scale organization and processes that were once hidden. For instance, genes have been shown to interact in highly modular ways, a finding that should foster future drug discovery. Networks provide a powerful framework to investigate relationships between biological entities, and how these entities interact to produce emergent behavior. To understand the workings of things like the immune system, we’ll need to model and simulate network behavior.

Some of the more mature network research in biology focuses on network motifs. Motifs are small patterns that occur much more frequently than one would expect if the design of the network were random. Recent research has found motifs that are shared across genetic regulatory networks, neural networks, and the even the Internet. The surprising fact that some motifs are shared across different networks suggests that similar processes shape these networks. The structure of these networks is far from random, and scientists are trying to identify design principles that underlie the evolution of these networks. Some systems biologists are trying to understand the function of various motifs in regulatory networks, such as feed-forward and bifan motifs. Check out this paper on network motifs, which also happens to be the most cited ecology paper in the last decade [1].

My work this summer at the Tang Lab focused the design and organization cell-cycle networks. We know that a huge number of genes and proteins interact to orchestrate the cell cycle, and we know the role of some of these regulators on specific parts of the cell cycle. But we lack a coherent account of how the cell cycle as a whole emerges from these interactions. Understanding this requires uncovering the relationship between network structure and function. In this case, the network function is the cell cycle (which roughly corresponds to an activation cascade).

Working with a network model of the cell cycle, I’ve examined the relationship between the architecture of the network and its behavior. One technique I’ve used is to computationally search for all the networks that are capable of performing cell-cycle function, in hopes of observing structural regularities. This kind of work may help us understand the design principles of biological networks that have exhibit definable behavior like the cell cycle. Intriguingly, similar approaches have been harnessed to study the relationship between amino acid sequence and the three-dimensional protein structure.

What design features make some networks robust and others less so? This is a challenging and fascinating question that applies to many kinds of networks, like banking networks and the Internet. While biological systems have evolved to be robust, we only have a superficial understanding of the structural features of networks that endow these systems with robustness. Many researchers believe that we can apply network principles from biology to human-made systems, like the financial system, to protect against crises.

While networks seem to be everywhere now, it is important to remember that a network is simply a mathematical abstraction, a roadmap of relationships. But as representations of biological systems, networks are bridging the theoretical and experimental realms in biology. They should be at the forefront of biology for years to come.


If you’ve read this far you’ll probably like this video on network science and big data. The video features physicist Albert-lászló Barabási, who believes that networks will revolutionize drug discovery.


[1] Fox, Jeremy. “And the Most Cited Ecology Paper Published in the Last 10 Years Is…” Dynamic Ecology. <;.

A Melange of Nationalities

By Vivian Wan

Although summer is over, I’ve had these responses gathering dust on my hard drive for a while, and I really wanted to showcase the wonderful people I worked with over the summer. The Hansel Lab is not only diverse in that we had a wide range of positions from undergraduate to postdoc, but also in that almost everybody was from a different country — and the few Americans in the group were of different backgrounds. I, for example, am Chinese-American.

It was natural for us to be curious about each others’ cultures, reflecting upon our similarities and differences over lunch. In particular, it was always interesting to hear about non-American education systems in addition to American college experiences outside of UChicago. So to compile all these different points of views into one location, I interviewed the members of the Hansel Lab, asking them about their pathway through science.

1. Tell me about your education and scientific background, especially in contrast to that offered in the US. 
Christian, PI (Germany): For high school, I went to a monastery school where students picked four subjects to specialize in for the last three years of high school. I picked German (similar to English literature classes), Biology, History, and one other subject. My impression is that Germany offers a stronger education in high school, where students stay for five years, since more was taught. Whereas in the US, it seems that the emphasis and the brunt of education is given to universities.
The difference between Germany and the US is that students proceed directly to university — as in there is no undergrad — which is 5 years long. You can think of it as undergrad and med school combined. Your 5 years are split in half. The first part is where you take general classes, so there is a wide variety of subjects from zoology to botany, but there is little to no laboratory work although students may have summer lab jobs. The second half is where you focus on research. I personally spent my first 2.5 years at the university specializing in zoology, and I spent the second 2.5 studying human brain anatomy in Zurich, Switzerland. There, I worked in labs.
The second option after high school is to go to specialized schools. The translation from German is “specific high school.” If you wanted to become, for example an engineer, you would attend these schools. They are like technical schools but at a higher level.
Tahra, PhD candidate (US): I went to a small private, all-girls school for high school so there weren’t a lot of options for science classes. We did have some APs, but there were only two of us in AP physics which is what I took. At Cornell, it was all about engineering all the time. We had tons of mechanics, physics and design classes, but only bioengineers had to take bio at all. We had to determine our major by our second year and must have completed general bio by the end of that year as well. For BME (biomechanical engineering), we also had to take biochem and at least two other upper level bios, but that was about it. I got into neurobiology from my research experiences, not class.
Giorgio, Postdoc (Italy): My interest in science made me choose the so called “Liceo Scientifico” (scientific high school). In the Italian system, within each type of high school, classes are determined nationally, with little or no choice at all, apart from some optional classes usually in the afternoon with no tests or requirements, such as drama or competitive sport teams. This had 5 years of math, 3 years of physics, 2 of biology, 1 of chemistry, 1 of “astronomical geography.” These classes were compulsory together with humanities and physical education. This is considered “scientific” especially because, differently than the “classical” high school (“Liceo Classico”) it requires much more math and physics and does not requires ancient Greek. However the scientific method experiences with an experimental approach is generally very limited, and, in my case, including only some physics experiments and few simple biological experiences such as DNA extraction.
After high school, still looking forward to the idea of basic science to enjoy, understand and take care of the environment or the human organism, I chose a program in Cellular and Molecular Biology. The program required traditional general classes together with more specific ones focused on biochemistry and cellular/molecular biology. It included, for the 3 years BSc, courses of basic science, genetics and molecular biology, evolutionary and environmental biology, animal biology.
After my BSc I choose to complete the MSc in the same program. The course work was pretty intense but there was probably much less lab classes than in American programs, offering only some small lab classes, a quarter in a lab at the third year and a full year in a lab at the second year of the MSc. The program, as often for science in Italian universities, was not very flexible as it may be here, with just few optional courses, and also less focused, probably, on pragmatical issues rather than in acquiring a broad and sound scientific culture.
Heather, Postdoc (Canada): I went to the University of Western Ontario for my undergrad. I decided in my first year that I wanted to do science, and took a variety of general science courses. In my second year I focused in biology, and took more bio related stuff (organic chem, biochem, ecology, ect.). In my 3rd year I applied and was accepted into a medical science program and decided to take physiology among other things (pharmacology, immunology, anatomy). In my 4th year I was in the honors physiology program and took mostly advanced physiology courses.  However, the university also required us to take courses in languages and another social science or humanities. I chose to take English in my first year, and psychology throughout.

At the end of it all, I graduated with an honors in bachelor of medical sciences (H.BMSc) with a specialization in physiology and a major in psychology.

I did my graduate school at the University of Toronto. The first 18 months I was in a Master’s program, and then I defended and passed a transfer exam, and switched to a PhD program. After a year in the PhD I passed another exam (similar to the “quals” they have here but we don’t call it that). I was in the department of physiology, but I was also in the program of neuroscience so the few courses that I had to take in grad school were all neuroscience related. I defended my PhD thesis after 4 years.

2. What research opportunities were open to you?

Christian, PI (Germany): There were many opportunities to work in labs during the school year and the summer, however there were no established fellowship programs.
Tahra, PhD candidate (US): I worked in labs both during the school year and in the summers (as a tech, no fellowships). There were plenty of research ops if you actively looked for them and many times research could be counted as a class for credit.
Giorgio, Postdoc (Italy): The lab experience was pretty limited to few very basic labs of microbiology, histology, physics, botany, biochemistry and molecular biology. However, as an improvement in comparison to the system running in the previous decade, we were required to spend a quarter full time in a lab with our own small project at the third year of the B.Sc. At the end of the M.Sc., however, a full year was dedicated to work full time on an experimental project. The most common way to have a study abroad experience was to enroll in the European Exchange “Erasmus program”. However my personal choice was to find on my own a summer internship in Germany, unfortunately with little support from the university and professors, if any. Another opportunity I got during my PhD (although not specifically from my program) was an international summer school in Neuroscience in Japan, where I could work in a lab for a couple of months.
Heather, Postdoc (Canada): Most undergrads don’t start working in labs until their 3rd or 4th years. There aren’t a lot of fellowships for undergrads so most were volunteers or paid summer students. I didn’t start working in a lab until my 4th year honors thesis project. Then I became a paid summer student in the same lab that summer.
3. What kind of research have you done in the past?
Christian, PI (Germany): I started working on the cerebellum (what the Hansel Lab studies today) from the start, but studied the visual cortex for my PhD. In the cerebellum, I studied nitric oxide signaling. It was significant as it was the first gas transmitter found, a part of white matter pathway.
Tahra, PhD candidate (US): Until this lab, I worked in a physiology lab doing data analysis rather than acquisition. I wrote computer programs and reconstructed simulations of actual neurons to use in an ongoing study of ALS. I also worked in a neurobiology lab with zebrafish doing imaging experiments and helping to create a software that could track and analyze the fish’s movements throughout the night for studies on circadian rhythms.
Giorgio, Postdoc (Italy): I have started my experience in neurobiology during my internship in Germany and, soon after during my M.Sc project, during which I had my first experience about neuronal transmission and plasticity. Since my PhD I worked on the neuronal plasticity, first focusing on structural plasticity of cerebellar climbing fibres and their regenerative properties (using in vivo gene silencing and himmunohistochemistry), then moving to an experimental model for multiple sclerosis (the experimental autoimmune encephalomyelitis) in collaboration with people with a variety of expertise and looking both to cerebellum and striatum. Finally I have been recently working here at the UoC on synaptic and non-synaptic plasticity of cerebellar Purkinje cells mainly by means of patch-clamp.
Heather, Postdoc (Canada): Most of my research was focused on the cerebellum. My undergrad research was on the arm movements/mechanics involved with throwing a baseball. I was interested to see the differences with increasing skill and the deficits involved with patients with cerebellar lesions or disorders.  My graduate research was on motor learning in the cerebellum. I studied eye movements in cats, and looked at a the vestibulo-ocular reflex, which a reflex whose learning is dependent on synaptic plasticity within the cerebellum. I also studied whether mGluR1 and GABAB receptors were involved with learning in this reflex. In general I did a lot of behavioral stuff before I came to Chicago.
4. Any other observations on schooling?
Christian, PI (Germany): University in Germany are like state schools here with large student bodies. There are no dorms, and students find housing in apartments. American universities let students have more liberty in choosing subjects & study matter.
Heather, Postdoc (Canada): I think schools in Canada and the USA are pretty similar except for the following:
– There is a distinct difference between colleges and universities in Canada. Colleges are like community colleges and offer diplomas. Only universities can give degrees.
– Canadian universities don’t require the GRE or any other standardized test for admissions. We apply based on our high school marks and courses.
– There are not a lot of private universities in Canada. There are none in Ontario (where I am from), everything is public.
– Tuition is a lot lower. Ontario has the highest tuition rate in Canada, and I paid about $4,000-5,000/year during my undergrad. My last year of grad school the tuition as about $7,000/year. Right now, there are riots and protests in Quebec because students think they are paying too much for tuition. They only pay about $3,000 per/year.

Second Week Event Bulletin

The Chicago Biological Investigator is part of a group that is collaborating to promote the common interests and goals of science and health-related student organizations and students majoring in various sciences at the University of Chicago. Last year, the group (HS RSOs) held an informational panel on the various RSOs involved in winter quarter and a science game night in the spring. Part of what we’re around to do is have a way in which to let people know about what relevant events are happening, so we hope to have a weekly post with an updated list of upcoming events.

Wednesday, October 8

The Triple Helix Fall Social: 6:00 pm – 7:00 pm, Bartlett Trophy Lounge

Friday, October 12

Fall Career Fair [UCIHP]: 12:00 noon – 4:00 pm, Ida Noyes Hall

Tuesday, October 16

Duke University School of Medicine [UCIHP]: 12:00 noon – 1:30 pm, UCIHP Suite

Wednesday, October 17th

12:00 noon – 1:15pm, UChicago Medical Center H-103

“Emerging Controversies in Organ Transplantation”
Robert Veatch, PhD
Georgetown University

Trivia night for 1st and 2nd years: free pizza and prizes [UCIHP Fellows Event]: 6 pm, BSLC 115

Tuesday, October 23

The Triple Helix’s Higgs Boson Lecture: 6:00 pm – 7:00 pm, BSLC 115

Wednesday, October 31

UCIHP Trick or Treating: “What’s Tricking You?” Tell us and earn a treat!: 3:00 pm, UCIHP

Further Shenanigans in the Murine Cerebellum

By Vivian Wan

My work over the past weeks has been focused on studying the possible influence of the duplicated chromosomal region (in our autistic mouse model) on cerebellar Purkinje Cell degeneration or development issues.

Golgi stained murine cerebellum sections.

After a long wait for the results of our Golgi staining trials, we finally were able to see the fruits of our efforts under the microscope. As shown in the picture, we sectioned the Golgi stained cerebelli into thick slices and mounted them on glass slides. To analyze images of Purkinje Cells from these sections, I use a program called ImageJ. Specifically, I use ImageJ plugins NeuronJ and Scholl Analysis for tracing of dendrites and for branching analysis [1], respectively. The former, NeuronJ, detects the path of a dendrite as you trace it, making it easier for tracing as opposed to tracing freehand. From these tracings, I have been gathering data on the sum of the dendritic arbor lengths.

I am also quantifying branching of the dendritic arbor. One way to do this is with the second plugin, Sholl Analysis, which quantifies intersections of dendrites within a certain area [2]. The program analyzes the number of intersections within concentric circles of increasing diameter from a chosen point — in our case the cell body — and outputs a plot, showing the number of branches enclosed within the ring-like area as a function of the distance from the chosen point. The second way to measure branching requires more particular definitions of dendrite branches. For Purkinje Cells, there are proximal dendrites, which, in addition to other functional characteristics, have sparse numbers of spines, and distal dendrites, which exhibit a high density of spines. For this second method, I categorized branching by orders: the first order were distal dendrites branching from the proximal dendrites.
10x microscopic view of a Golgi stained Purkinje Cell.

10x microscopic view of a Golgi stained Purkinje Cell.

I also analyzed the Purkinje Cells in other criteria, such as measuring the distance from the start of the dendritic arbor to the end and measuring the sum of the length of the proximal dendrite. We are also considering looking at spine density (from highly magnified images) as a measurement.

On a different note, we also obtained more promising data on protein expression levels from our autistic mouse model. As exciting as it is to find a direction in which to pursue more research, as it often goes, new results only lead to more questions. Challenges may be, for instance, apparent contradictions with previous research, more controls may be necessary in light of the new results, or the new results may even challenge the original hypothesis, demanding new interpretations. How can the problems be resolved? In our case, we will continue to explore our mouse model by looking at our data with more methods. Unfortunately, as I have come to the end of my fellowship, the questions will be left open on my end. But I am thrilled to have contributed to this project and await further findings by others.


[1] Grasselli et al., PLoS One. 2011;6(6):e20791.

[2] Sholl DA, J Anat. 1953 Oct;87(4):387-406.


Mission: Crystallization Status: Success!

By Jen Hu

While my last blog post was a bit depressing, I have new and exciting news: I finally saw crystal growth. It was glorious.

Though the crystals, small and delicate, cannot be diffracted by x-ray due to their small size and shape, I am still extremely proud! I’ve grown some pretty, but currently useless, snowflake-like crystals after nearly 10 weeks of research.

As a reminder, my project involves the crystallization of a high affinity FTO protein mutant together with a single-strand DNA oligomer. While it is exciting to see crystal growth, there are two big questions that must be asked:

  1. Do the crystals truly contain a DNA-protein complex?
  2. Does our high-affinity mutant retain the same demethylation activity as FTO?

In addressing the first, there is no definitive way of ensuring that the crystals contain both DNA and protein until they can be diffracted. However, it is comforting to note that our control crystallization plate from several weeks ago, containing only protein, did not contain any crystals under the same conditions.

In addressing the second, we would like to use HPLC (high performance liquid chromatography) analysis to investigate the activity of the double mutant. The main goal of the HPLC is to analyze the DNA after the FTO-DNA reaction: is m6A still present or not? We have tried once but as this is a new technique, it may take a little while the master of the HPLC. Running an HPLC is quite similar to running a gel filtration column. A UV absorbance is used to track what flows out of the column, this time set at 266 nm for observing the presence of nucleic acids. As before, I will be watching carefully for peaks to appear, corresponding to nucleotides. We should see either four peaks or five, depending on if m6A is present or not.

In the meantime, as I refine my HPLC analysis, we’ll be optimizing the crystallization buffers for even better growth. We’ve adjusted the salt concentration, pH, and PEG size but the optimized crystals, while larger than the first set, are still needle clusters and too thin to diffract. This means more crystallization screens but with renewed hope! I think I’ve now tested over a thousand different crystallization conditions…

Life’s a Great Balancing Act

By Jen Hu

“So be sure when you step.
Step with care and great tact
and remember that Life’s
a Great Balancing Act.
Just never forget to be dexterous and deft.
And never mix up your right foot with your left.
-Dr. Seuss’ “Oh the places you’ll go!”

Over the last few weeks of purification, remembering that “…Life’s a Great Balancing Act,” may be the only thing preserving my sanity. My project this summer is to crystallize a complex of the FTO (fat mass and obesity-associated) protein with a single strand oligomer of DNA. However, even before the crystallization process starts, one must isolate the protein. This is easier said than done.

From start to finish, each minute counts in purifying your protein. However, freshness and purity often work at odds with each other. One could have a fresher protein that has undergone less purification or an older protein that is extremely pure. There are three types of purifications that I have been working with: Nickel Affinity, Mono-Q, Gel Filtration.

Quick description of each:

To use the Nickel affinity column, your protein must have a histidine tag that was incorporated onto it during its insertion into the plasmid. The supernatant from the lysed bacteria is passed through the column and any protein with a His-tag sticks to the sides. This column provides the greatest specificity and is the fastest to run. It is my favorite column for a reason.

The Mono-Q column separates based on charge. Each protein has a pI, or isoelectric point, which represents the pH at which it has a neutral charge. Therefore, by using the buffer of an appropriate pH above or below the pI, one can separate proteins based on charge. Unfortunately, our column is somewhat broken and runs very slowly. On top of that, ordering scientific equipment in China takes longer than it should. Yikes.

The Gel Filtration column separates based on size. The larger the protein, the faster it flows through the column because there are fewer routes it can take through the gel. This is the most satisfying column because you can watch the UV absorbance reading and see your protein eluting out. I literally hold my breath and watch for the peak to emerge. It’s a bit funny the things that we live for in the world of science: a band on a gel, a peak on a graph, or a p-value less than 0.05. The UV peak(s) that I live for:

The first peak is likely a dimer of the protein of interest and therefore would elute out more quickly than the monomer, represented by the second peak.

As for optimizing the purification process, the main issue I’ve been debating is whether to deal with the problematic Mono-Q column or just let it go. In the end, I decided to let it go and replace the normal protocol of Nickel, Mono-Q, Gel Filtration with Nickel, Gel Filtration, Nickel. This optimization has given me the best purity and allows me to run the quick and specific Nickel column twice.

Optimization is no easy task. When you think about it, our entire lives are based on optimizations of happiness, health, and other factors. I’ll be honest: sometimes, the protein purification process becomes a question of whether I want to stay in the lab until 10pm or explore the city of Beijing. In Chicago, I don’t think I would mind staying late since I have all year to explore the city. However, opportunities such as working in a foreign country are hard to come by. As well, it doesn’t help that even a pure and fresh protein does not guarantee a good crystal: there are hundreds of screening conditions to do afterwards. The path of science is a long and arduous one, because if you think about it: there is no end. There is always more that can be done. However, I think we should strive to keep our lives balanced and keep working towards those optimizations, not relegate ourselves to accepting a life shut indoors without sunlight or proper nutrition. Optimizing the purification has not only given me a better protein, but also a better quality of life. For me, a better quality of life includes trying all the delicious food Beijing has to offer, such as jellyfish appetizers!

Exploring the Cell Cycle: Models as Maps

By Michael Begun

Jorge Luis Borges’ oft quoted “On Exactitude in Science,” written to resemble a literary forgery, goes like this:

“…In that Empire, the Art of Cartography attained such Perfection that the map of a single Province occupied the entirety of a City, and the map of the Empire, the entirety of a Province. In time, those Unconscionable Maps no longer satisfied, and the Cartographers Guilds struck a Map of the Empire whose size was that of the Empire, and which coincided point for point with it. The following Generations, who were not so fond of the Study of Cartography as their Forebears had been, saw that that vast Map was Useless, and not without some Pitilessness was it, that they delivered it up to the Inclemencies of Sun and Winters.

Suárez Miranda, Viajes de varones prudentes, Libro IV, Cap. XLV, Lérida, 1658″ [1]

Maps can be incredibly useful, but as Borges humorously illustrates, they grow useless with increasing complexity. While we usually think of a map as a way to navigate a city or landscape, they also help us make new discoveries. Fault lines, for example, can be uncovered by mapping earthquake locations. In 1855, English physician John Snow mapped London’s cholera cases, demonstrating the outbreak’s mode of transmission (a pump on Broad Street). Snow’s work debunked the belief that cholera was transmitted through the air [2]. Both of these maps helped uncloak phenomena by illustrating relationships between the right information (geographic location and earthquakes or cholera cases). Obviously, the usefulness of maps like these depends on elegantly representing important information while omitting irrelevant stuff.

You might be wondering why I’m talking about maps — I think a map is a good metaphor for many computational models used in systems biology. Both maps and computational models allow us to work with simplified representations of the world, and well-designed ones reveal previously unnoticed patterns or processes. The principles of map-making serve us well when designing models of biological phenomena. For example, Borges’ passage echoes a key precept of modeling — use the simplest possible model that can answer your question. I touched on the cell cycle in my first post. Here I’ll discuss some dynamical models that biologists use to study the cell cycle, focusing on Boolean networks.

Because our knowledge of biological systems is often quite limited, we can use models (like maps) to compile known information and relationships, which sometimes stimulates new discoveries.  Computational models in biology attempt to simulate key regulators of the cell cycle, groups of neurons, or signal transduction pathways, to name a few examples. A well-designed model can shift our understanding of biological systems. For example, in The yeast cell-cycle network is robustly designed, Li et al. showed that biological networks are uniquely structured to be robust against mutations and environmental variation. They did this by building a simple model (a Boolean network) of yeast cell-cycle regulation, then doing some computational analysis.

Biologists often draw pictures of protein pathways, depicting interactions with arrows that signify activation and inhibition. A major goal of systems/computational biology is to animate these pathways — to make them jump off the page, so to speak. Instead of viewing the pathways statically, we want to see how they behave dynamically as a product of their interactions. Computational modeling of regulatory pathways involves turning our knowledge of regulatory interactions into dynamic form. We can then run simulations of regulatory pathways, and use models to analyze the pathway or even perform biological experiments that we are unable to do with actual cells.

The landscape surrounding the Ming Dynasty Tombs in a light drizzle, about an hour northwest of Beijing.

When systems biologists model the cell cycle, they try to represent the levels of activation of the key regulators like cyclins. There are many ways to do this, but the most common methods include ordinary differential equations (ODE) or Boolean networks. In an ODE model, the activation of regulators are typically represented using Michaelis-Menten kinetics. Since ODE models can calculate the states of activation to a precise degree, they are sometimes called quantitative models. A key downside of these models is that we need to know the parameters for the differential equations, which must be determined experimentally. In an ideal world, we would know all of a system’s components, interactions, and kinetic parameters. But usually we don’t, so it is nearly impossible to build ODE models for large networks. ODE models are suited to small systems with known parameters, and are usually not feasible for larger systems, say with hundreds of interacting genes.

Enter Boolean networks. If ODE models are 10-megapixel photographs, Boolean networks are Monet’s water lilies (apologies for the analogy). They provide a way to investigate the dynamics of regulatory networks in a coarse-grained fashion. Boolean networks are sometimes called “qualitative” models because they cannot represent activation or expression levels precisely. Yet they are incredibly useful, and have become an invaluable tool of systems biology.

In a Boolean network, nodes represent proteins or genes, and can be in one of two states: on or off. While protein activation is obviously not this black and white, this binary simplification is well-justified for larger networks that display qualitative behavior (like the cell cycle). It also makes the computation a lot more tractable. The dynamics of Boolean networks are defined by the interactions between nodes. When we link all these interactions and nodes together, we get a network whose state evolves over time. Time is discrete in Boolean networks; we simply update the state of the network by applying Boolean rules.

This summer I’m working primarily with Boolean networks, which were first introduced in theoretical biology by Stuart Kauffman. While Boolean networks were initially a purely theoretical model of regulatory networks, they are capable of accurately modeling real genetic interaction networks. For example, in The yeast cell-cycle network is robustly designed, the authors designed a model of known interactions, then simulated it. The network evolved in a way that matched the real, biological cell cycle (the network’s nodes cycled through a series of states that corresponded to the biological cell cycle). This model represents a kind of map for the real yeast cell cycle, and has inspired much experimental and theoretical research.

Boolean models are also maps in a mathematical sense, because they directly link structure (network topology) to dynamics. This is very powerful for examining the relationship between network structure and function. In particular, this capability lends itself to investigating the connection between structure and robustness. Much of the research I’m doing this summer involves looking for patterns in networks that impart dynamical robustness, the topic of my next post.

Computational modeling is quickly becoming a mainstream method in biology, complementing traditional experimental approaches. These models allow us to investigate questions that we’ve been unable to explore using traditional experimental techniques. Through my experience at the Tang Lab, I’m finding that systems biology represents an exciting nexus of experimental approaches and computational modeling. These approaches are becoming increasingly integrated. The Tang Lab, for instance, has computational as well as wet lab facilities. The graduate students come from diverse backgrounds, with the most common being physics, biology, and computer science. This integration of experimental and computational approaches promises to stimulate biological discovery. We can, after all, explore the territory much better with a map.

Check out these papers for more on modeling in systems biology:

Modeling and analysis of gene regulatory networks

Defining Network Topologies that Can Achieve Biochemical Adaptation

Boolean network models of cellular regulation: prospects and limitations


[1] Borges, Jorge Luis. Jorges Luis Borges, Collected Fictions. P. 325. Trans. Andrew Hurley. Penguin, 1998.

Accessed via:

[2] Brody, Howard, Michael Rip, Peter Vinten-Johansen, Nigel Paneth, and Stephen Rachman. “Map-making and Myth-making in Broad Street: The London Cholera Epidemic, 1854.” The Lancet 356.9223 (2000): 64-68.

Coarse-graining (Part II)

By Grant Rotskoff

I’ve been working at the Kungliga Tekniska högskolan in Stockholm for a little over a month, so I thought I’d say something about the distinctive working habits in Sweden. I can’t speak for wet-lab scientists, because we’re a theoretical group, but above all of our servers, project machines, and GPU clusters, the most important piece of equipment is certainly the espresso machine. Swedes take coffee very seriously and it’s rarely drunk passively. Having coffee or “fika” (which is a verb and noun describing the ritual) punctuates the day with pleasant social gatherings and caffeine pulses that keep everyone in a positive mood. In addition to excessive coffee consumption (both the Swedish and Danish members of the group have, perhaps, five cups per day, which I am incapable of handling), lunch is traditionally the day’s main meal, here. Most large workplaces have super-high-quality cafeterias. Providing well-priced, healthy, fully-accessorized meals, nearly everyone eats at this sort of establishment. Nationwide, Thursday is pancake day. Food, both groceries and at restaurants, ranges from over-priced to unaffordable, so taking advantage of lunch is a must.

Kungliga Djurgården, literally, the Royal Game Park, a popular weekend picnic locale.

Fueled by espresso and lingonberry pancakes, I’ve been hard at work on Markov modeling techniques. Currently, I’m attempting to use Markov models, which I described in detail in my previous post, to provide information about the free energy of protein dynamics. A good Markov model describes the dynamics of a protein in full, so the free energy of transitions between states of the system can be perfectly described. However, a good Markov model requires a tremendous amount of sampling. In practice, this means doing a large number of short simulations of a protein, all which have different starting states. We don’t, however, know how to choose the starting states for these simulations in an optimal way. The problem I’m working on attempts to convert an undersampled Markov model into a protocol for additional sampling which will help elucidate the pathway between states of the system. The above description is a bit abstract, but we’re applying the method to explore the native state dynamics of a few familiar proteins.

Day to day computational biophysics isn’t always so abstract. The bulk of my work involves programming, as implementation and verification of the method that I’ve devised is just as important as its description. I’m working within a framework called Copernicus, which is a python library that describes itself as Folding@Home for supercomputers. It was written almost entirely by Sander Pronk, who is a researcher in the group. In addition, the Folding@Home people contribute to Copernicus. Programming in this framework is exciting; it’s still in a very early, flexible stage but it’s an extremely powerful tool. We also spend a lot of time working through the biological implications of our models—searching for the chemical root of structural changes, working out timescales of dynamical behavior, and hoping to grasp the low-level functionality of nature’s molecular machinery. This, in the end, is what we’re after.

A Symposium at Peking University

By Jen Hu

Hello all! Jen Hu here. As you may know from Michael Begun’s post, I am one of the four students researching in Beijing this summer. I am currently working in Professor Chuan He’s lab at Peking University (he also has a lab at UChicago). My main project this summer will be to crystallize a high-affinity mutant FTO (fat mass and obesity-associated) protein in a complex with a DNA oligomer containing its most likely physiological substrate, m6A (methylated adenosine). This protein’s most likely role is to demethylate m6A in RNA. It is of particular interest because it was the first protein discovered that linked a physiological phenomenon to a reversible RNA modification, suggesting that there might be a layer of genetic regulation at the level of RNA.

While this is all very interesting, I’d like to devote my first post to the recent symposium that was hosted by Synthetic and Functional Biomolecules Center (SFBC) of Peking University’s College of Chemistry and Molecular Engineering. The symposium, titled “Frontiers at the Chemistry-Biology Interface” brought together chemists and biologists from around the world. In case you were wondering, the talks were given in English as almost all Chinese college students are proficient in it. In many ways, this symposium represented my research experience in Beijing so far and I’d like to spend some time on what stuck out in particular.

At the end of the very first presentation, the traditional question and answer portion began. However, just as hands were being raised, the host stated that priority would be given to the students for asking questions, not the visiting professors. I think I was most startled by this because I had an expectation that schooling in China cultured a mass-test-taking mentality and less independent thinking. This assumption has been based on the gaokou, which can be thought of as the standardized test of all standardized tests. It is a two to three day test which just about decides where you’ll go to college. It happens once. That sort of extreme pressure leads many high school seniors to spend their senior year studying how to succeed on the test, not necessarily learning at the same time. Because of this, I thought it was great to encourage students to engage with the talks (undergraduate and graduate students alike)!

Personally, my favorite of the talks was by Professor Matthew Francis of UC Berkeley. His talk covered several of his interests but I was most interested in his research on viral capsid DNA aptamer conjugates. In drug delivery research, there has been much interest in moving away from the use of antibodies for targeting specificity. The reason for this is that antibodies often are riddled with problems of immune-recognition. The Francis group published a paper in 2009 about an efficient way to couple such DNA aptamers to the surface of viral capsid carriers (which contain the drugs). Up to 60 DNA strands may be attached (41 nucleotides long)  and bind to a tyrosine kinase receptor on Jurkat T cells. I found the idea of using DNA selectivity to elegantly simple. After talking with other students in my lab, it seemed that this talk was one of their favorites as well. I think part of the reason was that because English is their second language, an elegant idea such as this translates much easier.

At the end of the talks, the biggest thing that struck me was the ambition of the Chinese scientific community. Both in the opening and closing remarks, the speakers excitedly stated their hopes to see this symposium become one of the most prestigious symposiums in the world. China is expanding on nearly all frontiers, especially science, and the symposium really exemplified that ambition. This is not specific to Peking University; after talking with an undergraduate in my lab, I learned that there are plans to construct two new colleges geared specifically towards math and science by the Chinese Academy of Sciences. As well, because there are so many incentives to be a college student, there is an ever increasing supply of driven and eager students. This is especially true at Beida and Tsinghua, where tuition, housing, and dining are heavily subsidized by the government. Eating in the dining hall at Beida for the last few weeks has cost me less than 10 rmb per meal most days (about $1.60 US). Nearly no one moves out of the student dorms because they are far more affordable than anything in Beijing. China’s ambition has been noted in many other industries but its ambition in education has been understated. Working here has given me just a glimpse at the ambition of the Chinese scientific community.