From Jan. 12-14, the EuroSyStem Workshop in Leipzig, Germany, will explore methods, data-management tools, and theoretical approaches in the area of computational stem-cell biology.
EuroSyStem is a venture launched in April 2008 to support stem-cell research and is funded by the European Union with members in academic institutions across Europe and companies.
BioInform spoke with one of the workshop’s organizers, mathematician and systems biologist Ingo Roeder, who is a research fellow at the University of Leipzig’s Institute for Medical Informatics, Statistics and Epidemiology. He leads a group there called DynaMo, for dynamic modeling of stem-cell organization.
Among his areas of research are mathematical models and computer simulations of stem-cell development and theoretical concepts of stem-cell organization. He did his PhD in theoretical biology at the University of Leipzig.
He applies bioinformatics approaches such as different types of mathematical modeling techniques including Monte Carlo simulations, stochastic processes, and differential equations to generate results that can deliver insight to experimentalists in this emerging field.
Roeder told BioInform about how the workshop will unfold and described some of the work in computational stem biology. The following text is a translated and edited version of the conversation.
What is this workshop going to cover and has there been one before this?
This is the first one and it is set up to show participants the range of computational methods in stem cell research. To some extent experimentalists and computational researchers are already collaborating and they are coming to the workshop to continue that work. Others are still getting oriented.
The workshop begins with an overview of computational methods with talks that reach beyond the scope of our research consortium and show participants the possibilities.
Then we scheduled a series of presentations on data analytics, for example, on genetical genomics by [stem cell biologist] Gerald de Haan and his group at the [University Medical Center of] Groningen in the Netherlands. Another aspect is databases, presented by bioinformaticist Simon Tomlinson from the University of Edinburgh’s [Institute for Stem Cell Research.] He and his group are building a database for stem cell information called StemDB.
So there already is a database?
It was developed in a previous project but must be re-worked because so many different new types of data have been developed. We need completely different functionalities than we needed five years ago.
For example, some groups are measuring RNA, others epigenetic markers, and a third something else. If you want to have a systems perspective you need a way to integrate all this data, to fit the information together in a big puzzle with some people working on parts of the sky, others on the field, and we still have to figure out what the big picture is, how to fit the different data types together and place them in a database so the community can access and use them.
The Dutch group, the group from Edinburgh, and our group are presenting methods that the consortium members now have at their disposal. It is kind of a tutorial and will have a theoretical and a practical component to let people try things out on computers.
The third part is about one specific kind of modeling, the modeling of transcription factor networks. It is about tissue stem cells and how we can model transcription factors that regulate gene expression. We [in Leipzig] work on hematopoietic stem cells, so it’s about how they regulate the way relatively primitive stem cells develop into so many functionally different types of cells.
We want to discuss, for example, how researchers in EuroSyStem can apply these models, decide what data types we need to validate these models. That part is about concrete examples of ongoing work in the project.
By the way, this is mainly an event for EuroSyStem members. Other participants are welcome; this is not a closed shop. Registration for this particular event is closed and there is, unfortunately, no way to accommodate people who stop by hoping to participate.
[ pagebreak ]
The scientists in EuroSyStem seem very diverse. What connects them all?
The link between all project members is stem cell biology. These days, also because new tools are emerging, we are all trying to understand stem cell science, for example, how does cell differentiation occur, from a systems perspective. Systems biology is becoming more popular with its use of mathematical methods to understand biological processes. And systems biology is one important aspect of this project.
The thinking within this project is about building a systems approach to stem cell biology, to offer background to the biology-oriented research groups and partners who are exploring the computational side of this science.
What role do companies play in this consortium?
The EU decided they wanted companies to participate, also because it will help if there are results that can be commercialized. Some companies, I believe, already have relationships with some of the academic groups or have developed a technology that has found applications in this research area.
Is the group thinking mainly about embryonic stem cells?
No, not at all. Embryonic stem cells are not at the center of the research and the much-discussed human embryonic stem cells are not really our subject either. We are mainly talking about adult stem cells or tissue stem cells in humans and otherwise embryonic stem cells in model organisms such as mice or fruit flies.
There are several research groups that work on cell modeling of pretty well-understood cells such as liver cells, and say even that is tricky. Since stem cells and fate choice are not nearly as well understood, how can you model them?
There are different modeling approaches, more simple ones and more complex ones. You can build a highly complex model, let’s say of an airplane like the Airbus A-380 or a Boeing plane, construct it, and simulate it mathematically in a computer with loads of details. And then the plane ends up flying.
If you want to study the physical reasons why a plane flies, how it generates lift, what drag is, all those basic but important principles, I don’t need the complex model. I can use a cardboard or plastic model of a plane to study lift and drag on that simple model.
The plane may not be the best example, but you can transpose that approach onto biology. In stem cells we don’t know many of the fundamental underlying principles, for example, ‘How are stem cells organized?’ We are trying to understand those principles with the help of models that are simple in the sense that they are very abstract.
For cells we know a lot about and for which we have plenty of measurements, we can try like with Virtual Cell [the University of Connecticut’s National Resource for Cell Analysis and Modeling] to put lots of details into the model. With stem cells it is a lot harder, because you don’t even know how to prospectively tell the difference between a stem cell and a normal cell. There are no biomarkers. And maybe it is the environment around the cell that makes it be a stem cell, these are fundamental and unanswered questions. So our approach is to start out with abstract models and add details to the models step by step.
Are there specific tools to do so?
We pick the computational tools that we need, there is no particular formalized language or programming tool that we specifically choose. … There is one example that we have worked on that shows how you can use simple models for practical applications possibly to help patients. But I don’t want to be cautious and not hype this.
We built a simple model of blood stem cells, applied it to chronic myeloid leukemia and published a paper in Nature Medicine in 2006.
We continued to work on this approach and in the workshop I will present how you can use this model to make predictions about certain new therapies. We haven’t described every detail of the stem cells in our model because they aren’t known. It described the general effect of a given drug therapy, which reduces cell division but doesn’t work on all cells. We have been able to show in the paper and further work that we can computationally describe the effect of the drug that is usually given to patients suffering from this disease.
That is one way of showing how far simple models can reach, such that they can offer a clinical angle.
Is the big challenge the dynamic nature of stem cells?
People are increasingly realizing that dynamics plays a crucial role. The idea that cells, let’s call them type A, eventually somehow develop into type B is not quite accurate. Cells can travel back and forth. A cell that often becomes a type B cell can, in a certain environment, become a type D cell.
There are dynamic ranges in cell development, this is my personal opinion. They don’t contain programming, they have patterns from which they can choose, and what they do depends on how they are influenced; it’s action and reaction. This is where computational scientist and experimentalists meet. You can make models all alone in your hideaway study but it’s meaningless without data. We need both: the data and the theory. That is the idea behind this consortium: bringing these two areas together.
Biologists might say they already know about the dynamic nature of cells, but often their experiments are not planned to take that into account. … Sometimes models help to give biologists another angle from which to look at experimental data. Of course finding out whether that is right or wrong will depend on experiments, but models can help find alternative ways of explaining data or looking at data.
Your own department institute in Leipzig has a name that includes medical informatics, statistics, and epidemiology. How does that relate to computational stem cell biology?
In Germany, there is a scientific discipline all about various quantitative methods in medicine, which used to be just statistics, then came epidemiology, then computational methods. At pretty much every German medical school, there are departments that focus on these methods and that explains the long name.
All of this is beginning to change as approaches like systems biology and bioinformatics play a role and institutes are moving from medical informatics, which is about imaging and electronic patient records, over to bioinformatics with computational approaches to all new kinds of data.
Our institute [at the University of Leipzig] include my research group called DynaMo but it originated with Markus Löffler who directs the institute and started working in dynamic modeling in biology 20 years ago, long before people talked about systems biology. That is why our institute became a little different, since he took over 15 years ago. I did my PhD here following this tradition, and have applied dynamic modeling to stem cell biology. … Actually not that many scientists applying systems biology methods in stem cell research, particularly not in Europe.
Recently I heard from Harvard Medical School researcher Walter Fontana that differential equations can be unpractical and problematic in modeling. Are they among the methods you are discussing at the workshop?
We do work with differential equations and he is right: You can’t do everything with them. They are the first things that come to mind when you mention dynamic modeling. But they are a way to describe groups of many cells and their average behavior. When we try to look at behaviors of single cells, it really is the question if differential equations are the right method or not. Here in our group we also use stochastic models and so-called agent-based models.
[ pagebreak ]
Those models let you look at single entities perceiving their environment and reacting to it. You set the rules of this interaction and you try to describe the rules mathematically. Agent-based systems are reality-oriented, not abstractions like differential equations that describe the average behavior, whatever that might be and which does not actually exist.
Agent-based systems try to represent the single cell mathematically as it receives signals and reacts to those inputs. So that means these systems aren’t as analytical, they are more like simulations and you compare the simulations with reality. That also explains why some mathematicians are not terribly fond of this method, because they say “Well you can simulate a special case, but you can’t find a generalized answer.” Biologists don’t mind that because they don’t want a generalized answer in the sense of a proof, but are seeking a flexible study method to study systems of interest and simulate them.
There are specific computer languages for agent-based systems. Is this workshop about building tools?
We have invited biologists, and a few theoreticians from the different research groups in the consortium to show them the spectrum of computational methods and approaches, tell them what is possible and where the limits of those methods are, so they know what they can expect from these modeling methods.
Many are not familiar with these methods; we only all began working together about six months ago.
Is part of the idea of the project to have models that you can hand over to stem cell biologists to download and use?
Personally, I think this is about experimental biologists and computational researchers working together. Models aren’t easy to use, it’s a bit of an art, just like following a lab protocol doesn’t assure that everything in the lab will run perfectly. In the future I think it might be best to build teams, with experimental and theoretical experts who have learned to communicate with each other.
We need to find a common language so mathematicians and experimentalists really understand each other. While having a pre-configured software tool can be helpful and it lets people play around with a few parameters, that is probably not the way to gain the most knowledge or to prepare a series of experiments. I think working together is the preferable way to move forward.