HIGHLIGHTS & CELEBRATIONS FROM THE 6TH ANNUAL RECOMB/ISCB CONFERENCE ON REGULATORY & SYSTEMS GENOMICS, WITH DREAM CHALLENGES

By Christopher M. Williams

In November 2013, close to 300 researchers gathered in Toronto, Canada for the 6th Annual RECOMB/ISCB Conference on Regulatory and Systems Genomics, with DREAM Challenges. In its second year as an official conference of the International Society for Computational Biology, the event featured 16 keynote lectures, 55 short talks, and nearly 100 poster presentations organized into two tracks that highlighted recent efforts to understand the organization, function, and regulation of complex molecular biological systems. In addition, the conference held the annual meeting of the Dialogue for Reverse Engineering Assessment and Methods (DREAM), which included short talks by the best performers in the DREAM8 Challenges.

Attendees from the Sixth Annual RECOMB/ISCB Conference on Regulatory and Systems Genomics, with DREAM Challenges held in Toronto, Canada

DREAM

In advance of this conference, DREAM posed three challenges to the systems and computational biology community that were designed to assess the relative usefulness of competing approaches for modeling complex biological systems, and to generate insights that could contribute to the solution of several practical biological problems. Over the course of the challenges, more than 600 teams registered and 369 submitted entries, doubling the participation of last year. The best performers in each of the challenges were announced at the conference, and representatives of the teams gave short talks explaining their approaches. These DREAM challenges were organized in close collaboration with Sage Bionetworks, and are the first round of challenges since DREAM and Sage Bionetworks announced their partnership in February 2013.

Fred Wright, Speaker, North Carolina State University

The HPN-DREAM Breast Cancer Network Inference Challenge, held in partnership with the Heritage Provider Network (HPN) and using previously unpublished data from a collection of breast cancer cell lines, invited participants to predict regulatory networks (subchallenge 1), generate dynamic models of phosphoprotein interactions (subchallenge 2), and devise novel approaches to visualizing these high-dimensional datasets as they change over time (subchallenge 3). The best performing teams in subchallenge 1 were Team DC_TDC (Artem Sokolov, Daniel Carlin, Evan Paull, Kiley Graim, Adrian Bivol, Chris Wong, Josh Stuart) from the Uniersity of California at Santa Cruz, and Team NMSUSongLab (Yang Zhang, Haizhou Wang, and Mingzhou Song), from New Mexico State University. Best performers in subchallenge 2 were Team GuanLab (Fan Zhu and Yuanfang Guan) from University of Michigan, Team StochasticChaos (Bahman Afsari, Ludmila Danilova, Alexander Favorov, Wai-shing Lee, Dane Taylor, and Elana Fertig) from Johns Hopkins University, and Team CGR (Ying Hu, Chunhua Yan, Chih-Hao Hsu, George Komatsoulis, Yu Liu, and Daoud Meerzaman) from the National Cancer Institute. The best performer of subchallenge 3 was Team ABCD (Wendy Hu, Alex Bisberg, Byron Long, Dave Noren, Amina Ann Qutub) from Rice University. All teams but one presented at the conference. For more information on this challenge, go to www.synapse.org/#!Synapse:syn1720047.

The NIEHS-NCATS-UNC DREAM Toxicogenetics Challenge, developed in collaboration with the National Institute of Environmental Health Sciences (NIEHS), the National Center for Advancing Translational Sciences (NCATS), and the University of North Carolina at Chapel Hill, asked researchers to participate in two subchallenges: 1) use genetics (SNPs), RNA-seq, and toxicity data to predict how individuals will respond to chemical toxicants, and 2) use intrinsic chemical properties to predict the toxicity in a population (median and interquartile distances) of new chemicals for which toxicologic studies have not been done. Team Yang_Lab (Tao Wang, Hao Tang, Xiaowei Zhan, Jichen Yang, Rui Zhong, Guanghua Xiao, and Yang Xie) from the University of Texas Southwestern) was the best performer in both subchallenges. Interestingly, this team was also the best performer in one of the two NCI-DREAM challenges of DREAM7 in 2012. They, as well as Team Cassis (Elsa Bernard, Erwan Scornet, Yunlong Jiao, Veronique Stoven, Thomas Walter, and Jean-Philippe Vert) from the Institut Curie, who were t

Left to right: Woychik Richard, National Institutes of Health; Elsa Bernard (Team Cassis, Mines ParisTech, France); Gustavo Stolovitzky, Conference Chair.

he second best performers in subchallenge 1, presented their work at the DREAM track of the conference. For more information on this challenge, go to www.synapse.org/#!Synapse:syn1761567.

The third challenge, which developed from a presentation by Markus Covert (Stanford University) at the 2012 RECOMB-DREAM conference, called on participants to explore a computational whole-cell model of M. genitalium (WholeCellViz) and identify a set of parameters that had been altered in comparison with those in a model wild-type strain. The challenge was designed to address what Covert described as one of the biggest problems he and his colleagues have faced; namely, the problem of parameter estimation in light of the vastness and inherent stochasticity in the model. The best performers in this challenge were Team Crux (Clemens Kreutz, Andreas Raue, Bernhard Steiert and Jens Timmer) from the University of Freiburg, who also preformed best in a similar challenge in DREAM7, and Team Whole-Sale Modelers (Alex H Williams and Jeremy Zucker) from Brandeis University). Both teams presented their work during the conference. For more information about this challenge, go to www.synapse.org/#!Synapse:syn1876068

Trey Ideker, Keynote Speaker (University of California, San Diego)

DREAM also featured two keynote talks. In the first, Trey Ideker (University of California, San Diego) introduced a new web-based database called the Network Extracted Ontology (NeXO), which integrates data from large collections of gene and protein interactions to build a hierarchy of cellular components and processes. This approach inverts the traditional use of ontologies in genomics, enabling a shift away from using ontologies (such as Gene Ontology) as prior knowledge for evaluating new data sets and toward using new data sets to construct and evaluate ontologies. Tim Hughes (University of Toronto) described some of his lab�s work with that of Quaid Morris to develop a database called the Catalog of Inferred Sequence Binding Preferences of RNA Binding Proteins. By identifying similarities between binding motifs across species, their approach makes it possible to predict motifs for 10,000 proteins. In particular, there seems to be a high degree of conservation in the 3� UTR, Hughes mentioned, suggesting that RBPs may play a major role in regulating the stability of transcripts.

Systems Biology

Beginning on the conference's second day, talks in the systems biology track covered topics related to the dissection of cell regulatory networks, studies of genetics and genomics, the characterization of signaling pathways and metabolic networks, the systems biology of development and disease, and recent efforts to clarify drug mechanisms of action. Keynote addresses in this track provided a compelling snapshot of some key challenges that the discipline currently seeks to address.

Peter Jackson (Stanford University) opened the track by describing how his lab uses network building tools to understand drivers of genetic disease. He focused on studies of the molecular mechanisms that are important in cilia, and showed how intersections between these networks and those involved in obesity and diabetes are providing insights into the underlying genetic and signaling networks that drive disease. Peter Sorger (Harvard Medical School) drew attention to the problems posed by the high variability of drug response in individual patients. He suggested that in the future, drug development will have to consider not just complexity in the regulatory networks that drugs target, but also the complexity of the perturbations that drugs cause. Bernhard Palsson (University of California, San Diego) discussed some recent applications of constraint-based modeling -- an approach that considers biological limits on possible network solutions -- for the analysis of metabolic networks. Palsson described some interesting findings that this approach has produced related to synthetic lethal drug targets, the characterization of enzymes within metabolic networks, and studies connecting genotype to phenotype.

Opening the second day of the systems biology track, Olivier Pourquie (University of Strasburg) showed how combining genomic approaches with developmental biology manipulations in Drosophila can provide insights into cross-talk between physiological systems. Levi Garraway (Dana Farber Cancer Institute) described work in his lab to identify a phenomenon called chromoplexy, in which multiple DNA translocations and deletions arise in a highly interdependent manner. He also discussed the difficulty of understanding drug resistance in cancer, suggesting the "disturbing possibility" that there may be a range of mechanisms through which resistance occurs. If this is the case, finding downstream targets where resistance pathways "coalesce" may offer a more effective strategy than targeting individual pathways, he argued. Brenda Andrews (University of Toronto) described a technology her lab has developed called the synthetic genetic array, which systematically creates double mutants in order to produce a comprehensive map of gene-gene interactions. She described some findings in her lab that have come from using this approach, as well as recent efforts to integrate high-throughput microscopy into this technology to observe protein abundance and localization within the cell.

The systems biology track also featured a special lecture in memory of Tony Pawson, who had agreed to participate in the conference as a keynote speaker before his untimely death in August 2013. Jeff Wrana (Mount Sinai Hospital, Canada), a former collaborator with Pawson, gave a moving survey of the importance of Pawson's insights into the nature of protein-protein interactions in signaling pathways, and of how his revolutionary discoveries changed the existing paradigm and laid the groundwork for later research on cell signaling.

Regulatory Genomics

Held on the final two days of the conference, the regulatory genomics track explored advances in scientists' efforts to understand genome regulation at multiple levels. Speakers presented talks about evolution at the molecular level, cell differentiation, enhancers of gene expression, genomic variation, and transcription factor binding, and about several new technologies for extracting biological insights from omics data. Collectively, the talks highlighted recent advances in the effort to explain the highly complex and dynamic molecular systems at work in cells.

Aviv Regev (Broad Institute) discussed a protocol her lab has developed for using single-cell RNA sequencing to study variability in dendritic cells. Experiments identified a small subset of "early responder" dendritic cells that initiate interferon signaling, activating an antiviral response and dampening the inflammatory response across a cell population. Ziv Bar-Joseph (Carnegie Mellon University) also highlighted the dynamic nature of regulatory networks, and described methods his lab has developed for combining genomic sequence, RNA-seq, and protein-protein interaction data with time-dependent data related to gene expression and epigenomic activity, in order to track changes in networks over time. Using an algorithm called mirDREM, his lab has reconstructed dynamic microRNA-regulated interaction networks to study lung development in mice, and ultimately hopes to use the approach to correlate underlying regulatory networks to disease phenotypes in patients.

Galit Lahav, Keynote Speaker, Harvard Medical School, Cambridge, United States

Galit Lahav (Harvard Medical School) presented recent work investigating the relationship between double-stranded DNA breaks and pulsing in expression of the transcription factor p53, a tumor suppressor that is mutated in 50% of all human cancers. In studies using gamma radiation and fluorescent reporters for quantifying DNA damage, her lab found that averaging the activation of p53 across a population of cells loses important features of the relationship between DNA breaks and p53 pulsation that can be seen more clearly in studies focusing on single cells. Describing studies of Drosophila development, Jason Lieb (Princeton University) explained that differences in the accessibility of enhancers of master regulator transcription factors on chromatin may account for the fact that different body parts can arise from the same genomic information.

On the final day of the conference, Emmanouil Dermitzakis (University of Geneva) explained how combining RNA-seq with whole genome sequence data offers new opportunities for understanding the regulation of gene expression by noncoding DNA. Surveying research his lab has conducted as a part of the SysCol consortium, he suggested that there may be germline DNA variants that do not themselves cause colorectal cancer, but become drivers of cancer via cis-regulatory effects once a cell becomes cancerous. Jussi Taipale (University of Helsinki) described his lab's efforts to undertake genome-wide analyses of binding affinities between DNA and transcription factors, including studies comparing human and Drosophila. Recently, they have begun using software called the Enhancer Element Locator (EEL), which aligns the order of transcription factor binding motifs. The researchers have identified 600 high-quality binding profiles for human, mouse, and fly genomes, and have gained valuable insights into the binding profiles of heterodimeric transcription factors. Brenton Graveley (University of Connecticut) gave an overview of his lab's efforts to characterize the dynamics and variability of alternative splicing events in Drosophila across time and in different tissues. In recent work focusing on the Drosophila Dscam gene, he has used a method called triple read sequencing to identify when and where individual isoforms of specific exons are expressed. His outcomes have uncovered the concept that alternative splicing is probabilistic, whereby the repertoire of RNA binding proteins expressed in subclasses of neurons determines the probability for each exon being expressed in that subclass of neurons.

For more information about this conference, go to www.iscb.org/recomb-regsysgen2013, where you can see a selection of talks and download the program guide, which includes abstracts for all of the talks and posters presented at the conference.

RECOMB/ISCB Regulatory and System Genomics Conference with DREAM 2014

The planning for the 2014 conference is well underway. Mark your calendars now and plan to join us in San Diego, California, USA, November 10-14. We are now accepting submissions of papers, abstracts, and responses to the DREAM Challenges. To learn more visit www.iscb.org/recomb-regsysgen2014-submissions/recomb-regsysgen2014-call.