This winter, I attended the fourth session of Winter School on Quantitative Systems Biology at International Centre for Theoretical Sciences (ICTS), Bangalore. The school is organized by the ICTS, Bangalore and ICTP, Italy in jointly as a part of their ICTP-ICTS Programme in Biology since 2012.
It was an extraordinarily fun-filled learning experience. I share below some impressions of the school hoping that it would be useful to some of you.
The winter school provides great learning experience and an excellent platform to make connections and get to know people. It has a strong theme yet remains broad enough to be of interest to an interdisciplinary audience. More importantly, you'd enjoy working in a mix company of people from different backgrounds, brainstorming and working together towards solving a problem of equal interest. And not to mention, it is a humbling experience to witness the diversity and scale of research being carried out by people around the world.
I heartily recommend Winter School on Quantitative Systems Biology to anyone who feels moderately excited with these prospects.
A comprehensive log of daily lectures, resources and notes are online. Also, check out the YouTube channel of ICTS for excellent video recordings of all QSB sessions.
Typically, each school has a central theme and accordingly distinguished speakers from around the world in the relevant fields are invited. Additionally, each session consists of an extra pre-school component where instructors introduce everyone to the necessary pre-requisite concepts for the main school. This turns out to be an effective strategy and significantly reduces the learning effort during the main school for the people from diverse backgrounds. Primarily because the main school is quite intensive with the series of lectures, research talks, poster presentation, and discussion sessions scheduled in succession. We even had a small project to do towards the end.
The official description of QSB2015 reads:
QSB2015 is centered around bacteria, the simplest known forms of life. The focus is on the basics of cellular life and the principles thereof. The emphasis is on the structures and processes that allow bacteria to survive, reproduce, evolve and form communities in their environment. This is a setting that is fertile for quantitative work, both theoretical and experimental, and for grasping some universals in biology.
To my surprise, most (~80%) of the school participants were either physicists or mathematicians. Being in a crowd of this composition, as a biologist (though my background is more on engineering side), provides you an excellent peek of how others perceive and understand the key concepts that we biologists often take for granted. For example, how to reduce the complexity of the system, formulate the problem, model the interactions etc., are a few key questions that are well understood in company of people who typically are better at it.
A gross categorization of topics along with speakers:
- Numbers in Biology (Mukund Thattai)
- Mathematical modeling of processes (Sandeep Krishna)
- Ecology and Evolution (Deepa Agashe)
- Primer on Biology (Supreet Saini)
- Metagenomics (Nagasuma Chandra)
- Bacterial cell physiology and growth
(Matthew Scott, Suckjoon Jun)
- Metabolism, regulation and response to the environment
- Structure and dynamics of biochemical networks
(Gary Stormo, Nathan Price, Nagasuma Chandra, Kunihiko Kaneko)
(Andreas Wagner, Matthias Heinemann, Tzachi Pilpel)
- Energy, information and computation in cells
Gary Stormo, one of the pioneering figures in Bioinformatics, gave several introductory lectures with gene regulatory networks at its core. A gene regulatory network is typically modeled using protein-protein and protein-DNA interactions, along with directed edges indicating positive (activation, association) or negative (repression, degradation) type of relation. This network consists of simple repeating components like feed-forward, auto-regulatory, coherent type connections, etc. These components are called network motifs. Although, it is easy to understand the behavior of a motif on an individual level, it is reasonably challenging to comprehend the complex emergent behavior of the total network they give rise to.
We may ask, how does a cell use these simple motifs to generate increasingly complex behaviors? Can we predict them? Even harder to answer, how to ascertain these explanations experimentally? A lot remains to be answered.
Additionally, I found the discussion of Savageau's Demand Theory and Riboswitches pretty fascinating.
I had a notion that a lot of work(*) has been done in mainstream microbiology causing people to shift away from it to modern buzzwords in biology. However, in this school, I realized that it still is the bread and butter for many scientists, and people are continuing to push the boundaries with their rich work.
Matthew Scott walked us through an entire theory of quantitative bacterial physiology using the classic case studies from the 60s. The key question being - How do the growth, size, cell cycle, macromolecular composition and morphology of a cell coordinate? Can we predict anything? In an answer to these questions, concepts of growth law, proteome partitioning, generation time versus division time puzzle, along with comprehensive quantitative models of growth were discussed.
Suckjoon Jun built upon the content discussed by Matt Scott and described his work on mother machine and how they have used it to conduct experiments. The key idea of his lectures were to infer the physiological and metabolic control parameters of a bacterial cell that can explain the phenomenon of size and growth homeostasis. They have proposed adder theory as an explanation to cell size homeostasis and hypothesize growth \( (\lambda) \) and size \((\Delta)\) as two independent and sufficient control parameters for bacterial physiological regulation. A great example of theory backed by rigid experiments. Thanks to Cooper, Monod, Helmstetter, Schaechter et al. in 50s.
Metabolism and biochemical networks
After regulatory network, metabolic network is the next most important towards realization of a whole-cell model (ultimate goal of systems biology?).
Nathan Price discussed some of standard techniques and algorithms used in metabolic network model construction and analysis. Apparently, creating a full genome scale metabolic network is significantly challenging and a humongous task typically involving data collection, literature mining, processing, analysis, validation, and curation as essential steps. At each step, there are tricky issues of dealing with missing values, and using right scoring function and database to use. Once you have reconstructed a network, next step is to find out its valid states. A widely used technique used for network reconstruction and analysis is COnstraint-Based Reconstruction and Analysis (COBRA) that employs physicochemical and biological constraints to limit the space of phenotypic states of a network in a given condition. Naturally, here also more data beats a complex model.
Once we have constructed a valid metabolic network, the next step could be to integrate it with the regulatory network. The enhanced network can now be used in retrospect to improve gene annotation, identify new metabolites and thereby improve itself. It also incorporates an amazing predictive power that may guide new biological hypothesis development. In fact, anything that you can do with a standard network can be done with this network, added with a biological significance.
On the other hand, Mathhias Heinemann took us to metabolism on a single cell level and discussed how cells regulate their metabolism. He proposed an elegant flux-sensing mechanism used by the cells to continuously measure their outside environment and carry out internal changes. A great study by his group demonstrated the phenotypic bistability of E.coli‘s central carbon metabolism and how it can be achieved by a simple use of few flux-sensors along the network.
Incidentally, the project component I was involved in was also based on metabolic network analysis. Perhaps, that should go in my next post.
Since the seminal work of Darwin and Wallace almost two centuries ago, evolution is still one of the hottest studied areas. The school witnessed several lectures and research talks on topics like evolutionary dynamics, molecular evolution with interesting theories put forward frequently.
At ETH Zurich, Andreas Wagner is trying to understand the origin of innovations, complexity and diversity of life using systems approach. The primary question being, “Are there any principles of innovability in nature that govern the arrival of fittest? Or is it just random mutation and selection over the evolutionary period?".
However, one must carefully distinguish innovation with adaptation to understand the question correctly.
His group is looking for answers by probing the metabolic phenotype and genotype space and their relation in a network. The analysis yields few curious results like -
- Two genotypes can differ (determined by a distance function D) strikingly despite having the same phenotype.
- A genotype can have quite a diverse neighborhood.
A consequence of these two properties is that it is really not that difficult for a genotype to acquire a new, unique phenotype. Combine with regulatory phenotype space, protein sequence space \((20^n)\), RNA sequence space \((4^n)\) etc., you have a multi-dimensional space to possibly generate any combination with very few step movements across the hyperspace. More info in his papers.
If this seems too computational to your taste, Tzachi Pilpel discussed how evolution occurs at the level of gene expression and related processes with a considerable experimental effort. A series of interesting papers on how transcription and translation networks may have evolved. The fact that even tRNA abundance could be optimized for, and can be used as an active means of control is beautiful.
Another exciting set of experiments were done by Tom Kuhlmann whose group wanted to understand the role of transposons in evolution. The experiments beautifully showed transposons in action, in real-time using a rather simple plasmid construct. Quite a few conclusions follow from these experiments that I encourage anyone to look up in his papers.
Energy, information and computation in cells
How do cells perform computation? Sensing the chemical gradient, Random walks, Slime molds etc. What is it that requires energy? Is everything inside a cell only driven by thermodynamics and a careful design of free energy cascade? Do these processes have any fundamental limitations that biological systems have to obey?
As it turns out, reality is not so simple.
Pankaj Mehta, a physicist from Boston University, works on elegant theories to answer these questions. They key motivation of his talks was to investigate no-go theories in biology, and their possible use in the design of synthetic cellular modules and circuits. He discussed two classical papers, Kinetic Proofreading by Hopfield et al. (1994), and Physics of Chemoreception by Burg and Purcell (1975). The two papers provide an excellent discussion of how cells employ active processes to increase their specificity. Turns out, that's where you break the time-reversal symmetry and hence have to consume energy.