 |
2016 ISCB
Accomplishment by a
Senior Scientist Award
Søren Brunak
|
2016 ISCB Accomplishment by a Senior Scientist Award: Søren Brunak
The International Society for Computational Biology (ISCB) recognizes an established scientist each year with the Accomplishment by a Senior Scientist Award for the significant contributions he or she has made to the field. This award is bestowed to scientists who have contributed to the advancement of computational biology and bioinformatics through their research,service, and education work. Professor Søren Brunak of the Novo Nordisk Foundation Center for Protein Research in Copenhagen, Denmark has been selected as the winner of the 2016 Accomplishment by a Senior Scientist Award.
The ISCB awards committee, chaired by Dr. Bonnie Berger of the Massachusetts Institute of Technology in the United States, selected Brunak as the 2016 winner. Brunak will receive this award and deliver a keynote address at the 2016Intelligent Systems for Molecular Biology meeting (ISMB 2016) being held in Orlando, Florida on July 8 - 12, 2016. ISMB is ISCB’s world class annual meeting that brings together computational biologists and interdisciplinary scientists from around the globe.
Brunak’s early interest in physics began with a childhood friendship with Jakob Bohr, grandson of Nobel Laureate physicist Niels Bohr. He considers this early informal exposure to physics instrumental in developing his interest in the field but acknowledges that his physics teacher in primary school also nurtured his interest. Brunak said, “I was primed by the fact that one of my childhood friends was Jakob Bohr. I grew up close to this family. Maybe I was therefore listening a little more to what the physics teacher would come up with. He was good at turning deep questions into something that could be understood by kids our age.”
Brunak went on to study physics formally as a graduate student but first took a detour in astronomy. He recalled, “First I went into astronomy, but I found it increasingly difficult to explain at dinner parties the importance of astronomy.” He then completed his Master of Science in physics in 1987 at the Niels Bohr Institute, University of Copenhagen. “I had been fascinated by computers. My masters thesis was titled The Physics of Computation1, and I studied what happens in the computer when it computes. I was inspired by the work of Rolf Landauer and Charles Bennett at IBM. They worked on determining if you could compute without dissipating heat in reversible physical processes where no information would be discarded.” It was Brunak’s interest in the work of Bennett that stimulated his interest in biology. “Bennett used DNA transcription as an example of how a computation(a copy operation) can be done without dissipating a lot of energy.My thesis was also about computation processes in the brain, which are related to machine learning. It’s also about throwing information away so what you are after is distilled out of the data. In the big data context, there is a huge information reduction need so my experience with the physics of computation has inspired me when designing machine learning algorithms that use a lot of information and end up with a yes or no, for example answering the question of whether a protein structure is helical or not at a given position in the 1 In Danish ”Computerens Fysik” amino acid sequence. A lot of bioinformatics is about throwing information away in a smart way so what you are really after is retained.”
Brunak completed his Ph.D. in computational biology in 1991 in the Department of Structural Properties of Materials at the Technical University of Denmark. He then went on in 1993 to become founder and director of the Center for Biological Sequence Analysis at the Technical University of Denmark, a large center that still exists. His early work in bioinformatics focused on protein structure. He recalled, “I worked with protein structure with machine learning approaches. Meetings were small,data sets were small. We tried to get a lot out of little. We were raised in the data-poor era. The machine learning approach is not only good for boiling down but also for extracting.” Even during this era of limited data, Brunak considered computer power an important priority.“During my early studies in the late 1970s I started with punch cards and huge magnetic tapes. During my PhD I obtained a grant for a fast four processor Apollo 10000 machine, and I later always spent a lot of money on supercomputers so computer speed was not a problem. Now it is a real problem because we have millions of instances of a genome. We are in a situation where computer science matters in a new way.I have been around computers so long so I’ve seen a lot of special purpose hardware developed. But people always go back again and again to the general purpose computer that can take any algorithm, or do things like align sequences with any setting.”
Brunak’s early bioinformatics studies looked at both structure and function and were not limited to sequence properties. Machine learning was integral to these studies, and he went on to write an authoritative text on the subject with Pierre Baldi in 1998, titled Bioinformatics: A Machine Learning Approach. Brunak developed several widely used algorithms rooted in machine learning including NetGene, which predicted introns and exons and splice sites, and SignalP, a signal peptide predictor. He recounted, “This was the time of the genome project, so we started doing exon and intron and splice site prediction using this method called NetGene. Both SignalP and NetGene were interesting in that they integrated several different predictors andexploited the same data from different angles. With NetGene, we had a splice site predictor and an exon predictor and we put them together and we got a much better algorithm out of it than staying just in just the splice site or coding/non-coding domain. In SignalP we also used the same data in two different ways.”
Brunak recalls some of the surprises of his early research. “My first Nature paper was a small paper in 1990. It was a paper where we predicted splice sites using machine learning with neural networks. We noticed a group of splice sites that the network really didn’t want to learn. We just kept training it and it still would not learn them. We started looking at them and it turned out that half of them were database errors, and the other half were more interesting, they were errors made by experimentalists when they interpreted their [sequence]gels. They had put the splice site in the wrong place. The would learn the rare, but true GC donor sites very late, but still learn them. It was an interesting paper that showed the power of machine learning--that it could be a little more clever than the quality of the data. Nature was getting tough on GenBank for removing errors, and here was a computational approach for cleaning up datasets. We used the same technique with SignalP to identify likely errors. [We thought] eithe rit’s an error or super unusual and therefore interesting. We could see in some databases, with signal peptides, that 10-15% of the data was wrong.” Brunak saw this tedious work as an important contribution to cleaning up data sets and spent several years on this effort.
During the Human Genome Project era, Brunak recognized with many others in the field the limits of gene prediction from sequence information alone. But his research using neural networks alluded to some of our present day understanding of the complexities of genomes.Brunak said, “It’s not surprising now that gene prediction was not100% successful. Now we know that there’s transcription everywhere and that what constitutes a gene is highly complex. In 1992 we had a paper in The Journal of Molecular Biology (JMB) examining the way show a neural network looks for gene features in order to produce a prediction. It turned out when we predicted introns and exons, it accepted into JMB, especially when you were trying to deconvolute
theoretically neural network parameters into some biological signal. The pattern it looked for was perhaps known to a referee as an early example of an enhancer. Part of the reason of the success of the machine learning approach is that we didn’t need to know upfront the features that were behind biological mechanisms.”
Brunak’s research focus has shifted direction in recent years during this era of large scale genome projects. In 2007, he was a co-founder of the Novo Nordisk Foundation Center for Protein Research at the University of Copenhagen. The Center’s main goal is to look for proteins of therapeutic value, and they are developing approaches that fit into a healthcare context. Brunak leads the translational disease systems biology group, which looks at genome, proteome and health data, where some cover the entire Danish population. Brunak explained, “I am interested in disease trajectories, the order in which you get diseases, comorbidities and follow-on diseases. If you get type 2 diabetes, you won’t get the same complications as your neighbor.
There are certain trajectories that are more probable than others.”For the entire Danish population, almost all personal information, including education, job status and health records, are tied to a Dane’s personal identification number. As such, researchers including Brunak have an abundance of unique data to work with, and much of his work has focused on boiling down this data into meaningful observations. “My contribution is to put patients into progression groups and interpret proteomics data. We for example group diabetics and will see how their trajectories differ. Having the ability to work from the molecular side and having health data is presumably going to be powerful. We have data from 11 million people living and dead. We also essentially have the family tree from the entire country because it’s encoded in the personal identification number.”
Brunak’s enduring contributions to computational biology and bioinformatics have spanned his career, and given the scope of his recent work, he is certain to make a lasting and valuable contribution to the field.
1 In Danish ”Computerens Fysik”
>> Return to List of Senior Scientist Recipients
 |
2017 Overton Prize Winner:
Christoph Bock
|
2017 ISCB Overton Prize: Christoph Bock
The International Society for Computational Biology (ISCB) each year recognizes the achievements of an early to mid-career scientist with the Overton Prize. This prize honors the untimely death of Dr. G. Christian Overton, an admired computational biologist and founding ISCB Board member. Winners of the Overton Prize are independent investigators who are in the early to middle phases of their careers and are selected because of their significant contributions to computational biology through research, teaching, and service.
ISCB is pleased to recognize Dr. Christoph Bock, Principal Investigator at the CeMM Research Center for Molecular Medicine of the Austrian Academy of Sciences in Vienna, Austria as the 2017 winner of the Overton Prize. Bock will be presenting a keynote presentation at the 2017 International Conference on Intelligent Systems for Molecular Biology/ European Conference on Computational Biology (ISMB/ECCB) in Prague, Czech Republic being held during July 21-25, 2017.
AT HOME IN THE EPIGENOME
Christoph Bock’s scientific curiosity was nurtured from a young age. His parents were math and science teachers, and while they did not push him to pursue these areas of study, he sees how this intellectually stimulating environment cultivated his natural curiosity and provided a critical foundation to his career as a scientist. Bock started exploring computer programming from the age of twelve, and he realizes in retrospect how learning to code was a valuable tool for practicing problem solving and scientific thinking.
During high school, Bock specialized in physics and math. His undergraduate studies at the University of Mannheim focused on computer science and business information systems, emphasizing machine learning and artificial intelligence. Toward the end of his studies, Bock yearned to tackle questions with broader relevance than the “toy problems” he encountered in his course work. Bock recalled, “Human biology seemed the biggest challenge and also most societally relevant. I was lucky that Jürgen Hesser offered a bioinformatics lecture and agreed to supervise my Master’s thesis at the University of Mannheim”. His Master’s research work focused on protein structure prediction and homology modeling.
Bock pursued his PhD studies in bioinformatics under the supervision of Thomas Lengauer at the Max Planck Institute for Informatics, studying epigenetic regulation of the genome. “Moving into bioinformatics and epigenetics, I had to catch up on a lot of important biological knowledge”, Bock recalled. “Reading papers and collaborating was key, but it also helped that my research focused on a field that was quite young, with ample opportunity to try out something new.”
He attributes much of his bioinformatics training to the time spent in the research group of Thomas Lengauer, and he has been grateful for his mentor’s continued support and collaboration throughout his early career. Bock also acknowledges the important guidance and feedback on his research provided by Jörn Walter, who co-supervised his PhD dissertation and introduced Bock to the international epigenetics community.
Bock’s first encounter with epigenetics data transformed his scientific career path, and he has been one of the first bioinformaticians that dedicated their work to epigenetic data. “When I started my PhD studies in 2004, the largest epigenetic dataset consisted of just over 100 data points, and one of my first papers established epigenome prediction as a means of inferring what was still very difficult and costly to measure experimentally.”
In the following years, next generation sequencing transformed the field, and it became possible to collect several billion data points in a single epigenome mapping experiment. This development created a strong demand for bioinformatic methods. “Working at the forefront of the epigenome revolution has been the highlight of my scientific research so far. But the most exciting times may still be ahead as epigenome research is starting to become broadly relevant for medicine, and I am looking forward to contributing to this development.”
Bock developed several software tools as part of his PhD, including BiQ Analyzer for processing DNA methylation data and EpiGRAPH for analyzing and predicting epigenome profiles in their genomic context. Bock went on to pursue postdoctoral studies under Alexander Meissner at the Broad Institute. There, Bock was exposed to the world of wet-lab biology, and he discovered the thrill and power of jointly developing new laboratory techniques and computational methods, which he used to study the epigenome of pluripotent and hematopoietic stem cells.
In 2012, Bock started his own research group at CeMM, an institute dedicated to advancing precision medicine through basic and translational research. He was hired by Giulio Superti-Furga, Scientific Director of CeMM, who, as Bock said, “Provided ample encouragement and let me try things that were initially quite far outside of my comfort zone, such as starting a wet lab and leading a next generation sequencing technology platform.” Bock has thrived at CeMM, where he has been able to work with many passionate researchers within the institute and at the neighboring Medical University of Vienna.
At CeMM, Bock has also developed his personal style of being a PI and mentor, acting as a catalyst of ideas and projects for an interdisciplinary team. He explained, “Our lab combines computational and wet-lab biology on roughly equal terms, with a good dose of technology development – including single-cell sequencing, CRISPR, epigenome editing, machine learning, and more. There is also an extensive network of collaborations, ranging from fundamental biology to immediate clinical applications in the area of personalized and precision medicine. It is a great privilege to work with such an interdisciplinary and creative group of smart people.”
Bock considers the success of his students and postdocs as a key measure of his achievement as a PI. He explained, “I work hard to maintain an environment in which every group member can build a great CV and learns what he or she needs to advance in their scientific career. So far, we have a 100% success rate of postdocs moving on to attractive PI jobs, which is great for young lab. But it is clear that helping others succeed in their career is not an easy task, and you need to create room for success and failure, and a safety net that encourages risk taking.”
Bock is still excited about epigenetics and what it can teach us about a cell’s past, present and future. He hopes that epigenomic data can be used to understand the regulatory logic of cells and to determine what goes awry in diseases like cancer. Bock said, “We are pursuing an engineering-inspired “build it to understand it” approach to cancer biology, where we combine CRISPR epigenome editing and computationally designed drug combinations to rationally reprogram normal cells into cancer cells and vice versa. Building upon a breakthrough technology for pooled CRISPR screening with single-cell sequencing, We seek to decipher complex biological pathways and gene regulatory networks in high throughput, in order to overcome the classical “one gene, one postdoc” paradigm of functional (epi-) genomics.”
Bock is deeply gratified to be honored with the Overton Prize, especially since he will receive his award this year in Prague. He said, “Ten years ago, I attended ISMB 2007 in Vienna – one of the first conferences where I presented my PhD project on epigenome prediction. That year, Eran Segal won the Overton Prize, and his keynote lecture about DNA’s regulatory code reinforced my interest in understanding the role of epigenome regulation in biology and medicine. ISMB 2007 was also my first time in Vienna, and the great impressions from that visit surely contributed to the fact that a job ad from Vienna caught my attention a few years later. This year, it will be my pleasure to give the Overton Prize lecture at ISMB 2017 in Prague, ten years and just a few hundred kilometers away from a truly career-defining ISMB 2007.”
top
>> Return to List of Overton Prize Recipients
 |
2017 ISCB Innovator Award
Aviv Regev
|
2017 ISCB Innovator Award: Aviv Regev
2017 marks the second year of the ISCB Innovator Award, which recognizes an ISCB scientist who is within two decades of having completed his or her graduate degree and has consistently made outstanding contributions to the field. The 2017 winner is Dr. Aviv Regev, Professor of Biology at the Massachusetts Institute of Technology (MIT), a Core Member and Chair of the Faculty of the Broad Institute of MIT and Harvard, and an HHMI Investigator. Regev will receive her award and deliver a keynote address during ISMB/ECCB 2017 in Prague, Czech Republic (July 21 - July 25, 2017).
AVIV REGEV: SEEING CELLS AS LIFE’S SMALLEST CIRCUITS
Aviv Regev first pursued her studies in a unique interdisciplinary program at Tel Aviv University, where she planned to focus on math and computer science1. But she discovered her interest in biology in the classroom of evolutionary biologist Eva Jablonka. Regev said, “I found biology because of her – in my first year as an undergrad, I took a genetics course with her in what is now called the ‘flipped classroom’ style. It was all abstract and inferential, and I was hooked.”
Before starting her PhD thesis at Tel Aviv University, Regev began to really think about cells as computers, particularly how they are comprised of circuits. Regev’s deep interest in this concept started at a conference where new approaches for modeling concurrent computation were featured, and she immediately considered this as a way to model cell circuitry. She was able to develop her ideas into a PhD project under the mentorship of Udi Shapiro and Eva Jablonka, and she recalled, “No one was working on this type of project. I did, however, have the great fortune to find Udi, who listened to my idea. He thought it was important. He didn’t want to work on it himself – but he wanted me to be able to work on it.”
Regev completed her PhD in 2002 and was selected to be a Bauer Fellow at the Center for Genomics Research at Harvard University, which gave her an intellectual community, as well as freedom and funding to build a small independent research group. She continued to pursue her interest in modeling cell circuits using gene expression and genomic data, and she developed with her colleagues several widely used algorithms and computational tools, including Module Networks and Synergy. She received early support from Andrew Murray at Harvard University, who shared Regev’s view that it was critical to deeply understand both theory and experiments.
In 2006, Regev was given a joint faculty appointment at MIT and the Broad Institute, and she started applying her cell circuit modeling algorithms to understanding different cell types, particularly cells of the immune system. Once again, Regev struck out on an independent line of research. She recalled, “Many people were not focused on circuits. But that was OK. I wanted to build and be part of a community that would open a new direction.” Eric Lander at Broad – a longtime supporter of female and young scientists with leadership potential -- stood behind and supported Regev’s independent scientific vision at this critical point in her career.
Regev’s independent research program has blossomed since she founded her lab, and she has applied her interest in how cellular circuits function and rewire to a wide range of biological questions, including how immune cells rapidly respond and differentiate, how hematopoietic stem cells develop into different blood cells, and how evolutionary changes occur over millions of years. She is both a computational biologist with keen instincts about how to extract insight from data, and an experimental biologist with the ability to create new methods and deploy cutting edge technology to address fundamental questions.
Regev continues to be drawn to seemingly intractable problems, such as biological scenarios with a massive number of hypothetical combinations or interactions, and making them into manageable problems by using sampling approaches. Her work on cells of the immune system reflects this focus, and she recalls one of her most unexpected findings emerged in 2012 while working with collaborators on applying single-cell RNA-seq to the analysis of dendritic cells. In contrast to present day technology, which enables the profiling of thousands of cells quickly and cheaply, this study only looked at 18 cells and required a tremendous effort. Regev recalled, “What we found was surprising in two ways. First, we were examining just one cell type which we thought was well-defined, so we did not expect to find major differences in gene expression between the cells – yet we saw 1,000-fold differences, from which we could recover regulatory molecules that accounted for this variation. Second, we discovered surprising patterns in alternative splicing – some cells preferentially used one isoform, others used another. We had been expecting the cells to use both. This added up to a bigger surprise: we weren’t really looking at one group of cells. We were looking at two subgroups, which we now know represent different developmental programs. A great deal of my work now focuses on understanding heterogeneity of this type – defining and understanding cells at a much higher resolution than we could before.”
Regev has passed along her love of science through her mentorship of postdocs, graduate students, and undergraduates, and outside of the lab she has maintained an intense teaching load and worked to overhaul the undergraduate genetics course to include quantitative content. She is grateful to her mentors who gave her freedom to pursue her own scientific interests and this has guided her style of mentorship. She said, “Today, when I see a person with an idea, I don’t care about career stage -- maybe they’re a grad student or an undergrad; maybe they are a seasoned staff scientist. I care about who they are. Do they show the seeds of independence, vision, and leadership? And what is their idea? If it’s challenging in entirely new ways, and can transform the world, it should be grown. As I mentor my students and postdocs, I try to let them spread their own wings – to be their colleague and collaborator.”
At Broad, Regev was recently appointed Chair of the Faculty, and in this role she has been focusing on initiatives to strengthen and build communities around computational biology and advance software engineering approaches to biological data analysis. She has served the greater computational biology community in many ways through work on numerous advisory boards, journal editorial boards, and program committees for conferences. Regev has been a reviewing editor for eLife since its inception, and more recently a senior editor with a major responsibility for computational biology, genomics and theory papers.
Regev is gratified by her selection for the 2017 ISCB Innovator Award, and she said, “Biology is such a data science now, and ISCB is the community that made that happen – so it is especially exciting and gratifying to be receiving such an honor from peers in this community.”
1 www.hhmi.org/scientists/aviv-regev
top
>> Return to Innovator Awards page
 |
2016 ISCB Innovator Award
Serafim Batzoglou
|
2016 ISCB Innovator Award: Serafim Batzoglou
2016 marks the awarding of the inaugural ISCB Innovator Award, which honors an ISCB scientist who is within two decades of receiving a graduate degree, and has consistently made outstanding contributions to the field and continues to forge new directions. The inaugural winner is Dr. Serafim Batzoglou, Professor in the Department of Computer Science at Stanford University. Batzoglou will receive his award and deliver a keynote address at ISMB 2016 in Orlando, Florida on July 12th, 2016 to mark this honor.
Batzoglou remembers having an early fascination with numbers and a sense of wonder about the universe around him as a young child. He recalled,“I was interested in math and science for as long as I can remember. Before going to preschool, I remember counting and adding large numbers, as well as wondering how big space is and where it ends.” Batzoglou’s curiosity was stoked by other cultural touchstones, including the novels of Jules Verne and Carl Sagan’s captivating Cosmos television series. These early experiences nurtured his interests in math, physics and computer science, and Batzoglou went on to earn two bachelor of science degrees in mathematics and computer science at the Massachusetts Institute of Technology (MIT). Batzoglou also encountered computational biology for the first time as an undergraduate. He said, “Upon entering college, I was deciding between two fields of study that had fascinated me during high school: astrophysics and artificial intelligence (AI). However, during the early nineties I felt that neither physics nor artificial intelligence was experiencing any exciting growth. At least that was my impression around 1995, when I had been admitted to a PhD program in computer science at MIT and was deciding on an area to focus on. I took Bonnie Berger and David Gifford’s class on computa-tional biology and felt that this was a research area with great potential where I could apply my computer science background to do science (rather than engineering).” Batzoglou credits his early mentors for giving him invaluable undergraduate research experiences, including Sorin Istrail, with whom he had a very enjoyable research summer during 1997, and his undergraduate research supervisor, David McAllester.
Batzoglou went on to earn his PhD in computer science at MIT under the mentorship of his advisor Bonnie Berger and co-mentor Eric Lander. During his early career, he was the lead algorithms designer and implementer of ARACHNE, one of the first programs for whole genome shotgun sequence assembly that was used for assembling several genomes including the mouse and dog genomes. Batzoglou’s early work included using comparative genomics for human gene identification. Together with his collaborators he developed multiple genome alignment tools, including LAGAN, and applied these tools to predict human genes from similar mouse genes. This work was significant to the emergence of the field of comparative genomics and its applications for identifying genes, regulatory regions, and evolutionary events, as well as other features across species.
Batzoglou and his lab focus currently on the development of algorithms, machine learning methods and systems for the analysis of genomic data. He recalls one of his more surprising research moments that stands out in his memory. “It was unexpectedly good news to me back in 2005 that Conditional Random Fields (CRFs) and similar flexible models that can learn very large parameter sets, could be so successful in a large variety of classic computational genomics problems including RNA secondary structure prediction, gene finding, and sequence alignment. My line of work on CRFs, together with my PhD students at the time, Chuong Do and Sam Gross, who led the effort, was when we began applying machine learning in earnest to genomics problems.”
Batzoglou aims to follow the example of his PhD mentor as he mentors students and post-docs in his lab. “I think I follow a similar style and philosophy as my PhD advisor, Bonnie Berger, in that I am supervising my students in a relatively hands-off, “first do no harm” manner, and to the extent possible allow them to define new research topics and directions.” He also looks forward to a time when more novel research can be supported through the grant system, and said, “The most talented and motivated students will do their best research when given freedom, encouragement and some resources. I would be supportive of a large fraction, say 25%, of the government funds to be earmarked for work on new directions – i.e., work on which the proposing PIs have no other funds and no related papers.”
Batzoglou’s novel research contributions have been recognized through several awards including being named among the Top 100 Young Technology Innovators in 2003 by MIT’s Technology Review Magazine and a 2004 NSF CAREER Award. His research publications alone also show his impact on the field, and his purely bioinformatics-based publications have been cited hundreds of times. Batzoglou has also served the computational biology community in numerous capacities, especially through his service as a member of the steering committee, program chair, session chair, and organizing committee member for various RECOMB and ISMB meetings.
Looking forward, he said, “The topic I am most fascinated by right now, although it hasn’t majorly influenced my research yet, is deep learning. Like many of my AI colleagues, I subscribe to the opinion that we are witnessing a major breakthrough in our ability to replicate (and improve on) a large fraction of the intellectual and perceptual capacity of humans. The victory of AlphaGo against Lee Sedol is a historic moment. From a personal perspective, I learned Go in 2003, and back then I considered it a midpoint in AI between where we were and full-blown human-level intellectual capacity (excluding emotions and human experiences, which AI hasn’t been focusing on as much). The significance of advances in AI cannot be overstated. I believe that AI will transform medicine, finance, construction, manufacturing, commuting and transport, and almost every other sector in society, over the next 20 years. I also believe that a large fraction of jobs in these fields will be made redundant. Re-education is great, but it is not clear at all what the new marketable human skills will be 20 years down the line. Perhaps anything involving human interaction, although that’s not clear. In terms of computational genomics and biomedicine, to the extent that we will be able to collect and agglomerate large genomic and biomedical datasets, application of AI will lead to breakthroughs that will start by vastly improving health care, agriculture and biotechnology, and continue to places that are hard to imagine today.”Batzoglou feels greatly honored to be selected as the inaugural winner of the ISCB Innovator Award, and said, “Innovation in computational biology – and in general – is largely a community process. I thank the committee for recognizing my work, and more importantly I thank my colleagues, mentors and foremost my students, with whom I should be sharing this Award.”
>> Return to Awards page
 |
2016 Overton Prize Winner:
Debora Marks
|
2016 ISCB Overton Prize: Debora Marks
The International Society for Computational Biology (ISCB) recognizes the achievements of an early- to mid-career scientist with the Overton Prize each year. The Overton Prize was established to honor the untimely loss of Dr. G. Christian Overton, a respected computational biologist and founding ISCB Board member. Winners of the Overton Prize are independent investigators in the early to middle phases of their careers who are selected because of their significant contributions to computational biology through research, teaching, and service.
ISCB is pleased to recognize Debora Marks, Assistant Professor of Systems Biology and director of the new the Raymond and Beverly Sackler Laboratory for Computational Biology at Harvard Medical School. She will accept this honor and present a keynote talk at ISMB 2016 in Orlando, Florida, on Sunday, July 10th.As a child and young adult, Marks never considered becoming any sort of scientist. She was fairly confident that she was either going to travel in time around the universe in the tardis with Dr. Who or be a professional political protester and save the world. However, math was a constant that captured Marks’s attention since she was a little girl and she recalls spending far too much time with math puzzle books, “math was one of the only activities that forced me to focus and calmed my brain”.
After school, Marks went to study medicine at the University of Bristol (in England) but left - now with some regret - after her 2nd MBChB degree as she was “more interested in theatre and politics than Latin names for bones”. Many years later, after babies and a variety of interesting jobs, Marks felt the pull back to academia and her first love, mathematics, and went on to complete an honors degree in mathematics at Manchester University. She recalled the focused scope of her mathematics studies during a time when students were not allowed to attend other courses, and interdisciplinary studies were not yet en vogue. “In England, when you did a math degree, it was a math degree and you were not supposed to dilute it with ‘lower value’ subjects like biology, computer science or even physics. I did however manage to attend an odd course on chaos and fractals that sparked my interest in the intersection of math and biology that continues to drive me today.”
Marks considers her introduction to computational biology somewhat unorthodox and recalled, “I came to computational biology by jumping in the deep end. After my math degree I won an award from the Wellcome Trust to research drug design for Leishmaniaand trypanosomiasis. I was given a Silicon Graphics machine and told to get to work. I hadn’t got a clue. I’d never used a computer. Because I had a math degree, they thought I was a computer scientist of sorts.”
In the wake of the Human Genome project, Marks went on to get a bioinformatics position at Harvard at a time when interest in the potential for computation in biological research intensified. MicroRNAs first captured her attention in mid-2000, and her work on these was eventually submitted as a PhD thesis at the Humboldt University in Berlin under the guidance of Reinhard Heinrich and, after Heinrich’s untimely passing, completed with Hanspeter Herzl as thesis mentor.
She recalled, “I accidently read an article in a biology journal lying around about what was then a semi-obscure discovery about small RNAs regulating development in worms. I couldn’t stop thinking about it. Do these little RNAs stick to more than one gene? Maybe humans have them?” MicroRNAs were obscure at the time, only two were known, not the category. The floodgates only opened after the discovery of tens (at the time) of the now named “microRNAs”, nearly identical in sequence across worms, flies and humans, published in three back-to-back papers in October 2001 and suggesting strong selection across many species. “So, what are they doing in all these organisms?” As more identified more and more of these microRNAs, it became obvious to Marks that a way needed a way to find out what processes they regulated. Unlike the much more complicated task of identifying targets of proteins, the chemistry of base-pairing suggested an obvious way to explore what microRNAs might stick to.”
Marks was the first, concurrently with the Cohen and Bartel labs, to publish genome-wide targeting by microRNAs, first in fly, then in human, having developed the miRanda algorithm that is still used today for target prediction. “Although many groups have now published papers on how to discover microRNA targets,” Marks said “At best, the science of microRNA target prediction is still imprecise, presenting an unmet challenge for the computational community”. These early papers highlighted the potential genomic scope of microRNA targeting across large pools of mRNAs and their many-to-many, cooperative and combinatorial regulation of protein expression, something we now take for granted. Struck by these indications of potential system-wide effects, Marks undertook work to investigate the function of small RNAs in the context of the cellular environment by using mathematical modeling, re-analysis of previously published experiments, and additional in vitro experiments.
She made several key findings that included demonstrating that mRNA half-life influences the effects of microRNA and siRNA targeting for thousands of gene targets and that mRNA and microRNA abundance impacts microRNA targeting (now thought of as the mRNA sponge effect, and, controversially, “ceRNAs”). She also showed that introducing siRNAs or microRNAs into cells results in attenuation of endogenously regulated genes. Marks explained, “This is a really important consideration for the interpretation of gene knock-down experiments and for therapeutic uses of small RNAs”. Her more recent work showed that for a given level of protein, adding microRNA regulation can reduce protein noise or fluctuations, especially for transcripts with low expression.
Quite by chance, Marks’s postdoctoral work shifted sharply away from microRNAs to the field of ab initio3D structure prediction of proteins. Together with Chris Sander, they revisited an older idea that had been advanced independently by the groups of Sander, Neher and Taylor in the mid-1990s of using covariation of residues in proteins across evolution to identify residues that might be in contact in 3D. They reasoned that if these inferred contacts were accurate enough, one should be able to fold a protein sequence using simple methods such as distance geometry and restrained molecular dynamics. The key advances were to use a statistically global model of covariation across the sequences that removes transitive correlations in the data, by using a probability model for entire proteins in the sequence family, not unlike a suggestion made by Gary Stormo and Alan Lapedes in 1999. She said, “We stumbled across statistical physics models that are used to determine inhomogeneous interactions in Ising models from observed data that contain transitive correlations. Listening to the team of Riccardo Zecchina, including Andrea Pagnani and Martin Weigt, in Torino, Italy made us think that their approaches to the analysis of correlated mutations could be important for the 3D structure-from-sequence problem. Working with their team in collaboration, I set off in the spring of 2010 to see if the maximum entropy method could work to find truly interacting co-evolved residues. If the computation was correct then co-evolved residue pairs should match contacts in known 3D structures, and they did nearly to the ceiling. I then cajoled a friend, Lucy Colwell who, like me, was also recent graduate, to join the project. What fun we had! Well, that is until it was time to try and publish it. Reviewers found it difficult to believe, but eventually we published the results at the end of 2011, and the EVfold community has grown ever since. Marks explained,
“The method is very democratic. It is fast and can be run on a laptop, even the folding, and relies only on gene sequences. Immediate applications were to proteins that are challenging experimentally such as large membrane proteins and to protein-protein complexes. More than five of the transmembrane proteins have since been crystallized and agree well with the predicted structures, such as the adiponectin receptor”. “A very effective mini-CASP,” she added with a smile.
After the protein folding breakthrough in the fall of 2010, implementation of similar methodology led her to the solution, published in 2016, of another hard and unsolved problem in computational biology, that of computing RNA 3D structures and of RNA-protein complexes just from sequence information.
Currently, Marks has a newly formed lab at Harvard Medical School and is building the Raymond and Beverly Sackler Laboratory for Computational Biology. She is continuing the “3D from sequence” work, including new types of biomolecular interactions and their conformational flexibility. The Marks lab is also going back to math and developing the core algorithms for sequences, and model inference for multidimensional biological data. At the same time, Marks is branching out with new applications that include the challenge of predicting the effects of genetic variation on disease risk and drug response, especially combinations of events, and particularly in antibiotic resistance “It may seem we are promiscuous in our choice of biological questions, but the underlying thread is one of solving problems that are hard to solve experimentally. One far-reaching question that I am increasingly less embarrassed to admit being interested in is: What makes us all different? With genomes in hand, surely we can now find out how much is nature, how much is nurture and how much is stochastic?”
Marks is grateful to her mentors and her “wonderful” Systems Biology department who have supported her throughout her unusual career path and feels greatly honored to be recognized with the Overton Prize. She is especially thankful to her new group members and her long-time scientific collaborator Chris Sander. She said, “I want to share the prize in spirit with all those who have tolerated and encouraged me despite the odds.”
Marks’s final message to young scientists (young in spirit, that is) is to “go big, go risky, and learn statistics (!)”
>> Return to List of Overton Prize Recipients
ISCB Innovator Award
The ISCB Innovator Award is given to a leading scientist, 10-20 years post-degree (or equivalent experience), who consistently makes outstanding contributions to the field of computational biology and continues to forge new directions.
ISCB recognizes that career paths may take many forms and that the definition of “early/mid-career” is fluid; supports researchers taking time off for maternity/paternity, care for a family member, an event of personal disability or other factors. A nominee may qualify for the Innovator Award even though their actual years since degree is above the set threshold. It is the responsibility of the nominator to indicate any time-off taken by the nominee when submitting the nomination form. ISCB may deduct the equivalent time for the maternity/paternity leave, care for a family member or personal disability from the set award thresholds using guidance established by the European Research Council (page 19, paragraph 5).
2025 ISCB Innovator Award: Fabian Theis
