Search

 

PressCNRS international magazine

Table of contents

Cover story 2/3

Computer science bolsters biology

Contemporary techniques in molecular and cellular biology are being used to decode DNA. This generates colossal amounts of information,1 which are then stored in databanks and databases. Similarly, a cell is the sum of innumerable molecular interactions which rely on networks of self-regulating processes that vary both in time and space in the cell. Now that these processes have been meticulously dissected, they must be reassembled in a way that allows us to understand how they function dynamically. This is biology's latest challenge.


“Biologists understand that they cannot succeed on their own,” states François Képès, cellular biologist and researcher at the Statistics and Genome Laboratory in Evry.2 Mathematics, physics, and computer sciences are better at dealing with large amounts of data related to complex and dynamic interactions. This explains the growing number of interdisciplinary partnerships now taking shape. From now on, equations, statistical analyses, algorithms, and simulations will all be part of the biological landscape. Using mathematical models is one of the most promising approaches to reconstructing the cell. The idea is to think of molecular interactions as belonging to a series of networks: a network of proteins, a network of proteins regulating gene expression, and a network of enzymes and metabolites. “Since these are complex networks, we begin by breaking them down into smaller modules and studying them at that level,” says Képès. “The objective is then to put them back together in order to understand the whole. We are assuming that the sub-networks have their own dynamics and, moreover, that they will maintain those dynamics when integrated into the whole.” Understanding the dynamics of small biological circuits is the work of Vincent Hakim and Paul François, two physicists at the Statistical Physics Laboratory at Ecole normale supérieure (ENS) in Paris.3 Instead of starting with a known circuit and trying to determine its function, the team is trying to build models of circuits for functions which are already known. “The novelty of our research is that we work with a collection of virtual cells,” says Hakim. Each cell is simulated on a computer using differential equations, and each evolves toward a desired function according to a succession of random mutations and circuit selections. “The models allow us to interpret what goes on in nature, but also let us use them to build artificial genetic circuits with novel functions, or even to recreate lost functions in, for example, a damaged cell,” the physicist explains. And that is not all: Mathematical modeling can propose new interactions and suggest new experiments, which in turn allow for the verification and improvement of the model itself.


Work in bioinformatics is headed in the same direction: Researchers have been focusing on the analysis of biological networks since the end of the 1990s, when complete genomes were first sequenced and large databases were created using DNA chips. “Of course, researchers have been modeling biological processes for twenty years, but it was a relatively marginal field as far as molecular and cellular biology were concerned,” says Vincent Schachter, director of the Bioinformatics department at Genoscope in Évry.4 “Now, theoretical advances will lead to qualitative models, which will be less detailed but better adapted to modeling global networks.” One of the difficulties confronting bioinformatics is the need to develop techniques that will simplify the interactions and circumvent the barrier of incomplete data while providing reliable and biologically relevant analyses. Indeed, the goal of global modeling is not just to provide a coherent view of our understanding of the dynamics of a network, but also to make predictions about how a network behaves. Results are already coming in: The first predictive models are now available for certain metabolic behaviors even at the level of the entire cell, for example for the non-pathogenic bacteria Acinetobacter ADP1.


Mathematical models allow researchers to synthesize data, to understand dynamics, to propose new experiments and research avenues, and, finally, to predict behavior. For François Képès, there is no doubt about it, “the best way to work is to combine experiments at the bench with the computer, to complement in vitro and in vivo experiments with those in silico.”
The biological community is increasingly convinced of a need to bring the different disciplines together in the interest of diversifying their approaches. A proof of this is in the growing number of institutions that are undertaking interdisciplinary approaches to biological problems, like the Systemoscope, an international interdisciplinary group dedicated to studying the complexity of biological systems. Another example is Helix (see above) or the Epigenomic program, whose co-director François Képès confides: “Above and beyond understanding biological processes, we hope to one day be able to manipulate them to find new therapeutic avenues.” Brighter and newer prospects await.


Stéphanie Belaud

 


Modeling  complex functions in living cells


Every biologist dreams of having a tool to simulate the complex functions of living cells. Helix1 is a team of about forty biologists and computer scientists who are working toward that objective in both Lyons and Grenoble. This group, a research powerhouse, is among the cream of the crop in European bioinformatics today.
The team originated out of a common interest in a highly interdisciplinary approach to molecular and cellular biology, computer science and applied mathematics.
“Our strength lies in our different backgrounds,” explains François Rechenmann, senior researcher at Inria and founder of Helix. “We are as capable of identifying biological problems as we are of finding the computational and mathematical techniques needed to solve those problems.” Indeed, biologists and computer scientists
are working together to develop, experiment with, and use a whole series of algorithms which help manage and analyze data obtained from different experimental techniques such as genome sequencing, DNA chips, or mass spectrometry.2


Modeling is a way of testing data against the understanding that biologists have of cellular processes. “Nevertheless, we don't focus on developing a single model that can exactly reproduce those processes, since such an illusory model would be just as complex as the cell itself,” cautions Rechenmann. Useful models must be reductionist in their approach since they tackle only those aspects that are relevant to explaining a specific scientific problem. “Hence we create several different models that are able to explain how cellular processes interact, processes formerly studied independently but which act together to produce the observed complex behavior,” Rechenmann explains. “If the simulations don't correspond to what we expect, then the model becomes interesting because it reveals what we don't understand and points out gaps and errors in how the model was formulated.” However, arguments of that sort are meaningful only if they remain part of an ongoing process of understanding.


“If a pharmaceutical company decides to simulate a cellular process in order to develop new drugs, the model will be of value only if its predictions are correct about the mode of action and potential side effects of a given molecule,” Rechenmann explains. For the time being, though, the use of modeling for pharmaceutical companies is not of great concern to Helix, as cellular modeling is generally considered by industry to be science fiction… but the fictional aspect of this science may soon disappear.


Fabrice Impériali

1. Helix (http://www-helix.inrialpes.fr/index.php) is a group of researchers belonging to Inria, CNRS, and Claude Bernard University, located both in Grenoble and Lyons. It is associated with the Swissprot group at the Institut Suisse de bio-informatique in Geneva and the Institute of mathematics and statistics at the University of São Paulo in Brazil.
2. Genostar Technologies, founded in 2004, sells the software and related services.

Contact
François Rechenmann
Francois.Rechenmann@inria.fr




IRI: joining forces is the future of biology

You are the founding member of an ambitious research institute, centered around the idea of interdisciplinary research. Why is it necessary today to put the life sciences into the hands of physicists and computer scientists?

Bernard Vandenbunder: Biology has always called upon other sciences. Chemists and physicists took an active part in deciphering the double helical structure of DNA and helped make molecular biology what it is today. To answer your question, it is more accurate to say that specialists from different disciplines are joining hands to tackle biological questions in a new way. Because most biological processes are the result of interactions among numerous components, a shift is needed from molecular to modular biology. Understanding the dynamic behavior of regulatory networks affected by multifactorial diseases is an essential step towards the development of new therapeutic strategies.

Is it difficult to collaborate with scientists from such different backgrounds?
BV:
The success of interdisciplinary projects relies upon people who are competent in their respective fields, and willing to take chances. It takes time to understand the perspective and jargon of another discipline. It's a demanding process, but rewarding for everyone involved. One of the benefits is the introduction of new concepts useful for understanding what life is.

At your institute, what will the other disciplines bring to the life sciences?
BV: The institute will be devoted to studying the structure, dynamics, and robustness of cellular regulatory networks. Every lab will be interdisciplinary, and the guiding principle will be a dialogue between theoreticians and experimentalists. For example, the elucidation of cellular networks will require development of new imaging techniques, new biosensors and the synthesis of new probes. The modeling and simulation of the networks will draw on different tools from mathematics, computer sciences, and theoretical physics. Finally, in order to manipulate the networks, we will make use of molecular biology, chemistry and micro and nanotechnologies. The
2000 m2 building which will be operational in 2007 will provide excellent working conditions for more than a hundred scientists: web and dry lab benches, lab support and areas for interactions.

Interview by Fabrice Impériali


1. CNRS senior scientist, Institut de biologie de Lille.
Contact: Bernard Vandenbunder,
bernard.vandenbunder@ibl.fr

Notes :

1. Thousands of genes have been decoded today and as many proteins, regulatory sequences for gene expression, biochemical reactions, metabolic pathways, and enzymes. 2. Laboratoire statistique et genome; Joint lab: CNRS / Université d'Évry.
3. Laboratoire de physique statistique; Joint lab: CNRS / ENS / Universités Paris-VI and VII.
4. Département de bioinformatique / Genoscope/ Evry.

Contacts :

Vincent Hakim, hakim@lps.ens.fr

François Képès
kepes@genopole.cnrs.fr

Vincent Schachter
vs@genoscope.cns.fr


Top

Back to homepageContactcredits