Message from ISCB International Society for Computational Biology Honours Gunnar von Heijne and Ziv Bar-Joseph with Top /Computational Biology Awards for 2012

Justin Mullins1, BJ Morrison McKay2* 1 Freelance Science Writer, London, United Kingdom, 2 International Society for Computational Biology (ISCB), University of California San Diego, La Jolla, California, United States

Gunnar von Heijne (Image 1) decided, tational methods.’’ Polypeptide energetics on whim, to brush up on his rusty, was only the start, however. schoolboy French. He took a few lessons By the early 1980s, molecular biologists and also subscribed to the French popular had begun to determine the sequence of science magazine La Recherche. amino acids in the signal peptides from Flicking through its pages, he came different proteins. However, little had across a short article on protein secretion been done to study the properties of signal Introduction and the signal hypothesis, the mechanism peptide sequences as a group. Each year, the International Society for that describes the way secretory proteins von Heijne changed this. He began Computational Biology (ISCB; http:// cross a membrane. comparing the sequences, looking for www.iscb.org) makes awards for excep- At the time, the late-1970s, very little recurring patterns that might help to tional achievement to two scientists. The was known about this process, but some identify them. ‘‘I looked at 20 to 30 signal ISCB Accomplishment by a Senior Sci- ideas were beginning to emerge. For peptides. Once you did that, some clear entist Award honours career achievement example, it was thought that a so-called patterns emerged that had not been seen in recognition of distinguished contribu- signal peptide—a short chain of amino before,’’ he says. tions over many years in research, acids—at the end of the protein carried He found that small, uncharged amino teaching, service, or any combination of the signal that determined how the acids tended to occupy certain positions in the three. In 2012 this award is going to proteins are transported out of the cell. signal peptide chains, the -3 and -1 Gunnar von Heijne of the Stockholm The article confused him, however. It positions. It is at this site that the signal University in . The Overton Prize showed a diagram of a hydrophobic signal peptide is later cleaved from the protein recognizes a young scientist in the early to peptide squeezing through the similarly once it has passed through a biomem- mid-stage of his or her career who has hydrophobic membrane. ‘‘That didn’t brane. This pattern has since become already achieved a significant and lasting make sense to me. The hydrophobic known as the (-3, -1)–rule. impact in the field of computational peptide ought to become anchored in the ‘‘Nowadays you would say this was a biology. In 2012, the Overton Prize is membrane,’’ he says. very trivial bioinformatics study,’’ he says being awarded to Ziv Bar-Joseph of The puzzle piqued his interest. He modestly. However, this was an important Carnegie Mellon University in Pittsburgh, solved it by calculating the energetics of discovery and von Heijne’s paper has since Pennsylvania, United States. a polypeptide chain passing through lipid become one of the most highly cited in the The recipients were chosen by the bilayer, which he published in 1979. This field. ISCB’s awards committee chaired by work by a theoretician created ripples in a He then used the newly discovered at the CNIO (Spanish field dominated by experimentalists. patterns to make predictions about pro- National Cancer Research Centre) in And so began the career for which he teins. For example, it became possible to Madrid. The winners will receive their now receives the Accomplishment by a create an algorithm that would take a awards at the ISCB’s annual Intelligent Senior Scientist Award from the Interna- protein sequence and predict whether it Systems for (ISMB) tional Society for Computational Biology had a signal peptide at the end. meeting, where they will deliver keynote (ISCB). ‘‘Gunnar is one of the big stars of Initially, that was not very useful. When talks. ISMB 2012 (http://www.iscb.org/ our field,’’ says , president molecular biologists sequenced a gene or ismb2012) marks the 20th anniversary of of the ISCB. ‘‘He is one of the few who messenger RNA, they generally knew the conference, and will take place July completely change the field using compu- what they were working on; whether it 15–17 in Long Beach, California, United States. Citation: Mullins J, Morrison McKay BJ (2012) International Society for Computational Biology Honours Gunnar von Heijne and Ziv Bar-Joseph with Top Bioinformatics/Computational Biology Awards for 2012. PLoS Comput 2012 ISCB Accomplishment by a Biol 8(5): e1002535. doi:10.1371/journal.pcbi.1002535 Senior Scientist Award: Gunnar Published May 31, 2012 Copyright: ß 2012 Mullins, Morrison McKay. This is an open-access article distributed under the terms of the von Heijne Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited. Perhaps it all began with the French lessons. As a young PhD student in Funding: JM was paid by ISCB to write this article. theoretical physics at the Royal Institute Competing Interests: The authors have declared that no competing interests exist. of Technology (KTH) in Stockholm, * E-mail: [email protected]

PLoS Computational Biology | www.ploscompbiol.org 1 May 2012 | Volume 8 | Issue 5 | e1002535 membrane proteins are the gates and gatekeepers to our cells; they determine what gets in and what stays out,’’ explains Rost. ‘‘That’s why around two-thirds of drugs target membrane proteins.’’ Understanding the structure of trans- membrane proteins provides crucial in- sight into how cells work and is also useful for future drug development. ‘‘That’s why the methods developed by Gunnar are so important,’’ says Rost. To continue his work, von Heijne set up the Stockholm Bioinformatics Centre at the beginning of the millennium. And today, von Heijne runs the Centre for Biomembrane Research in Stockholm, where he has brought together computa- tional, modelling, and experimental groups. Few places can boast the same breadth of experience under one roof. Throughout this time, von Heijne has maintained an impressive work–life balance Image 1. Gunnar von Heijne of . Photo by Max Brouwers. as a scientist, a husband, and a father. He says doi:10.1371/journal.pcbi.1002535.g001 that’s been possible, at least in part, because he was working in a new field with few would have a signal peptide on the end or either in or out. But which orientation competitors. ‘‘I never felt stressed that we’d not. should the protein take? be scooped. I work hard but not crazily.’’ But that changed when sequencing ‘‘We discovered a very simple principle Others clearly admire his positive ap- became faster and biologists started to that determines this,’’ he says. The regions proach, which he combines with a relaxed attitude. ‘‘He also looks ten years younger sequence things they didn’t know much that connect the transmembrane segments than he has any right to!’’ says one envious about. ‘‘The algorithms have continually contain positively charged amino acids, colleague. improved and are now extremely useful,’’ which give them an electric potential. The For a while in the 1980s, he spent half he says. simple principle is that the segments with the his time working as a science journalist for Secretory proteins have to move across greatest number of positive charges end up the Swedish National Radio. ‘‘You decide a lipid bilayer through a molecular inside the membrane, an idea that has since on Monday what you broadcast on Friday machine called a translocon. The signal become known as the ‘‘positive inside rule’’. so there is immediate feedback, which has peptide guides the ribosome that makes ‘‘This is very important work and a good pulse to it,’’ he says. the protein, towards the translocon. This provides some of the best data on But for von Heijne, doing science is triggers the opening of this protein-con- membrane proteins,’’ says Valencia, chair more satisfying than reporting it. ‘‘Radio ducting channel through the membrane. of the ISCB awards committee. stories have a short half life; they’re on air, But other types of protein only make the In the late 1980s, von Heijne began to then they’re gone,’’ he says. ‘‘The rewards journey partway, becoming embedded realise that he could gain significant in science are greater and longer lasting.’’ half in and half out of the membrane. insight into these and other problems by It’s surprising how far schoolboy French These so-called membrane proteins use doing experiments rather than just theory can take you. the same translocon machinery as the work. So he set up his own lab. ‘‘I trained secretory proteins. ‘‘So it was a natural as a chemist so I wasn’t a complete novice 2012 ISCB Overton Prize: Ziv step to start looking at these membrane in a wet lab,’’ he says. proteins next,’’ says von Heijne. This first idea was to see whether it was Bar-Joseph The part of the protein that ends up in possible to make proteins that inserted Ziv Bar-Joseph (Image 2) loves to run. the membrane is very different to the parts ‘‘upside-down’’ into the membrane. He He rises early and hits the streets and trails outside exposed to water. This transmem- could show that by changing the location around Pittsburgh where he lives, often in brane section must be much more hydro- of the positively charged amino acids in a training for a long-distance race. This phobic. So the trick to predicting which protein, it is possible to make it take up the dedication has paid off. He has the parts of a protein become embedded in the opposite orientation. enviable distinction of having run a sub- membrane is to look for the segments that This link between his theoretical and three hour marathon, a feat achieved by are most hydrophobic. practical work has been important for him. few amateurs. ‘‘Running is very important Once you know the transmembrane Bioinformatics studies often throw up to me,’’ he says. segments, an interesting problem is to patterns that may or may not have But it is not just in his running that Bar- determine how the protein becomes wo- biological relevance. ‘‘The only way to Joseph shows a willingness to go the ven into the membrane. For example, if it determine whether they are important is distance. As a computer scientist and has four hydrophobic sections, there are to do the experiments,’’ he says. computational biologist at Carnegie Mel- two ways in which it can be arranged in ‘‘It’s hard to overstate the significance of lon University in Pittsburgh, Bar-Joseph the membrane: with the termini pointing von Heijne’s work. Membranes and trans- shows a similar dedication as head of the

PLoS Computational Biology | www.ploscompbiol.org 2 May 2012 | Volume 8 | Issue 5 | e1002535 order to reconstruct the set of events over time and these have since been used in various other systems too.’’ Bar-Joseph has learnt to work closely with biologists who test the results. ‘‘If the algorithm predicts that ‘a’ controls ‘b’, for example, you can do the experiment to test whether that’s true.’’ That’s important because the patterns that the algorithms reveal must be biologically relevant. To better understand the challenges that experimentalists face, Bar-Joseph spent a sabbatical working in a wet lab doing exactly this kind of work. That taught him some valuable lessons. For example, wet lab work is not just a question of validating the model. ‘‘The results from the lab feed back into the model and enhance it. It’s a two-way street,’’ he says. Others have been impressed with Bar- Joseph’s approach to experimental work. Image 2. Ziv Bar-Joseph of Carnegie Mellon University. Photo courtesy of Carnegie Mellon ‘‘Ziv is an example of somebody coming University. from the theoretical side of things and doi:10.1371/journal.pcbi.1002535.g002 completely embracing the experimental approach,’’ says Rost, president of the ISCB. ‘‘It’s stunning how he is able to Systems Biology Group at the School of successes with a statistical approach called handle such a diverse set of technical Computer Science. ‘‘We have all been a hidden Markov model. That attracted methods.’’ impressed by the quality of his scientific Bar-Joseph who had studied this model. This process of feedback from biology to contribution and the novelty of the He also recognised that other earlier computer science has become an impor- approaches he has developed,’’ says Va- studies, attempting to reconstruct networks tant theme in Bar-Joseph’s work. One of lencia. in cells, were significantly limited: the data his recent successes is in explaining the Bar-Joseph gained a PhD in computer was a snapshot of a complex dynamic way fruit flies develop bristles on their science from the Massachusetts Institute of system but they treated it as if it were foreheads. These bristles are like aircraft Technology between 1999 and 2003. That static. sensors, measuring temperature, wind time turned out to be hugely significant, Clearly, biological systems change. speed, and so on. To work well, they need not least because computational biology ‘‘One thing I’ve been involved in is to be spaced in a very precise way. was undergoing a revolution. ‘‘For the first introducing dynamics into the algorithms The bristles grow from cells but clearly time, we were getting sequences for large so that they can cope with the way things only a small subset of cells. The cells do species. First, the fly, then humans. It was change in time. That requires different not know how many neighbours they have very inspiring,’’ he says. tools,’’ he says. or the local density of bristles nearby. So Initially, Bar-Joseph knew little about The approach has paid off when it what determines which cells grow into computational biology but took a class to comes to understanding regulatory net- bristles and the spacing between them? better understand the significance of these works and explaining how proteins control Bar-Joseph quickly realised that this advances and the problems they posed. ‘‘It each other. For example, yeast has about was similar to a problem that computer seemed to me that these types of problems 6,000 proteins. Of these, some 250 are scientists have wrestled with for 30 years. were well-suited for the machine learning control proteins and each of these, on This is the problem of determining the tools I had experience with,’’ he says. average, controls 100 or so other proteins. subset of computers in a network that One of the key problems was how to However, each control protein is itself control all the others. When each compare sequences either within species controlled by a handful of other proteins. computer in the network is connected or between them. Various researchers had Understanding a system like this is a to one computer in this subset (but no developed methods to do this using a tricky business. The static data can tell you two in the subset are connected to each branch of computer science called combi- what proteins control other proteins, but other), this subset is called maximally natorics, which essentially counts the that doesn’t tell you when and under what independent. number of similar patterns. conditions because that requires more Finding maximally independent sets is But while this works well when com- experiments. hard, particularly in large distributed paring two sequences, it’s not so good for Other types of data are more temporal networks. Computer scientists do it by comparing seven or eight sequences. It and can reveal how protein levels change assuming that every computer knows who doesn’t scale. Consequently, researchers over time. ‘‘The question we asked was all its neighbours are. began to experiment with probabilistic whether we can use this temporal data to Bar-Joseph realised that the fruit fly cells approaches that focus on the statistical try and recover the underlying network that eventually become bristles form a properties of the patterns. In particular, dynamics,’’ he says. ‘‘We came up with maximally independent set—they are con- computational biologists had significant methods to integrate these datasets in nected to all other cells but not to each

PLoS Computational Biology | www.ploscompbiol.org 3 May 2012 | Volume 8 | Issue 5 | e1002535 other. However, they do not know who The future holds many promising Additional Information their neighbours are and so must solve this problems for Bar-Joseph too. He is problem in a different way. His break- particularly interested in studying how The full conference agenda and regis- through was to work out how they did it pathogens interact with cells, how the tration information for ISMB 2012, in- and develop an algorithm that does the proteins from flu viruses interact with cell cluding details on when these ISCB award same thing while assuming no knowledge proteins, for example. ‘‘If we can recon- winners plus four other distinguished of the neighbours. ‘‘It takes a bit longer struct the networks of interactions then we keynote lecturers will be speaking, can be but that’s the trade-off,’’ he says. might be able to determine intervention found on the conference web site at This may have important applications points that will guide us to therapeutics,’’ http://www.iscb.org/ismb2012. The con- for wireless sensor networks that research- he says. ference will also feature parallel tracks for ers are using to monitor everything from He also wants to study the interaction proceedings of original research papers, ocean conditions to volcanic eruptions. networks in different species. Many of the highlights of recently published papers, ‘‘We only published at the beginning of genes in humans and mouse are similar, special sessions on emerging topics, late- 2011 so we don’t know if it will penetrate but drugs that work well in mouse often breaking research of peer-reviewed ab- the commercial world,’’ he says. don’t work in humans because the path- stract submissions, technology demonstra- Valencia is also impressed by Bar- ways, levels, and interactions are different. tions, and workshops presented by aca- Joseph’s broader contribution to the ‘‘We want to get more insight into this,’’ demic researchers, funding agency computational biology community. ‘‘He he says. representatives, and commercial vendors. is a member of the editorial board for the That’s clearly a long game. These are The conference also offers a commercial journal Bionformatics, so clearly his contri- problems that will require dedication, and non-profit vendor exhibition. butions go beyond this theoretical and talent, and endurance to solve. Exactly For a review of past ISCB award experimental work,’’ says Valencia. the kind of qualities you might find in a winners, please see http://www.iscb.org/ ‘‘That’s very good for a young scientist.’’ marathon runner. iscb-awards.

PLoS Computational Biology | www.ploscompbiol.org 4 May 2012 | Volume 8 | Issue 5 | e1002535 Copyright of PLoS Computational Biology is the property of Public Library of Science and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.