Bioinformatics and Systems Biology

Abstract

BioTechniquesVol. 46, No. 7 Tech NewsOpen AccessBioinformatics and Systems BiologyLynne LedermanLynne LedermanSearch for more papers by this authorPublished Online:25 Apr 2018https://doi.org/10.2144/000113177AboutSectionsPDF/EPUB ToolsAdd to favoritesDownload CitationsTrack Citations ShareShare onFacebookTwitterLinkedInReddit Challenging FieldsAccording to James E. Ferrell, Jr., Professor of Chemical and Systems Biology and Professor of Biochemistry at Stanford University in Stanford, CA, the field of systems biology can be looked at in two ways. One is to look at it on a large scale as bioinformatics, and the other is to look at it like a physicist, in a reductionist manner. “It might seem oxymoronic, but it's not. The old tools of reductionism can be used on a complex system,” he says. “Too often, the physics-oriented and bioinformatic-oriented approaches move forward as independent, non-overlapping arms. Something that connects the two will be important.” Tools to understand ‘omics data sets include statistical approaches, as well as non-linear dynamics from physics, which he characterizes as the reigning approach and a rich, wonderful field of mathematics.“The real challenge is to understand what's wrong in the cell and how to design better therapeutics.”Ferrell observes, “Part of the fun of systems biology is that we don't know how to move forward, so creative people will have a chance to make important contributions.” He himself is a systems biologist who studies biologic phenomena at levels that he describes as more complicated than the individual genes and proteins. The complex networks that result from the strategies nature has used in the evolution of biologic systems provide a challenge to Ferrell. His group is studying the system of regulatory proteins driving the cell cycle. “To me, one of the challenges now is to take these huge data sets or huge networks—that ‘omics people have been building and bioinformatics people have been collecting—and extract insight from them. The best way is to reduce huge networks to subnetworks and explore them in a reductionist fashion. It would be interesting to see if there are intermediate scales.” Examples of these large data sets are those gathered to understand regulatory networks in certain types of malignancies, in which certain proteins are characteristically up-regulated and others down-regulated. “The real challenge,” he says, “is to understand what's wrong in the cell and how to design better therapeutics.”Russ Miller using a large tiled-display wall to view real-time information on the grid. Photograph by Adam Koniak, Center for Computational Research, State University of New York at Buffalo, Buffalo, NY.More Power to YouRuss Miller holds appointments as Distinguished Professor of Computer Science and Engineering at the State University of New York at Buffalo and as Senior Scientist at the Hauptman-Woodward Medical Research Institute, in Buffalo, NY. He is leading the rollout of “Magic,” one of the most powerful computers in the state, which is expected to be used to provide solutions to previously unsolvable, computationally demanding problems in the areas of bioinformatics, computational chemistry, structural and systems biology, and medical imaging. Miller's Cyberinfrastructure Laboratory (MCIL) developed “Magic,” a system that integrates graphics processing units (GPUs) into a traditional rack of PC-based computers. These GPUs are high-end graphics processors typically reserved for computer gaming systems. As such, they are quite affordable, and also provide the opportunity to take advantage of their substantial processing power to solve certain scientific and engineering problems hundreds or thousands of times faster than traditional high-end clusters can. Furthermore, machines such as “Magic” are not only less expensive than a traditional cluster with the same computational power, but also require less space, cooling, and electricity. In fact, it is often the case that these machines require fewer staff to install and maintain the system.Miller says his group has been trying to leverage the cost-effectiveness of the latest new graphics processors—such as those produced by NVIDIA (NVIDIA Corporation, Santa Clara, CA)—which are often included in Intel-based PCs for those who will use the machines for advanced graphics and gaming applications. “As members of the high-end computing community, we often ask ourselves whether or not it is possible to take advantage of the incredible cost-effective computing systems that these graphics units provide. Many of us believe that GPUs can not only be used to produce important and interesting graphics, but to do great science and scholarship in a cost-effective manner. These graphics units are geared towards certain types of data, and data of a similar form are found in many areas of science, including molecular biology,” Miller explains.The “Magic” system is available to certified users who belong to established virtual organizations worldwide. In order to effectively utilize these systems, developers must either create computer programs that will run efficiently on them or port their existing computer programs to these new machines. Once a code is ported, it is typically available for general use on these architectures. “Porting codes to these machines is not difficult,” says Miller, “but it does require some effort. You can't just take code that was developed in the 1960s and run it on these new machines. Someone has to take code, e.g., the Smith-Waterman algorithm [used to search nucleotide or protein databases to find those with the best alignment] and port it to the new system.” Miller observes that some people enjoy this developmental process. “We live in a world where many important computer programs are available to the general community. People are willing to make their codes available so that others can take advantage of the advances made with such codes. Therefore, if the developer— who has a deep working knowledge of the computer program—is willing to port the code to these new architectures, it can then be made available worldwide.”“Data storage is a very real issue,” says Miller. “The world is capable of producing data so fast that we are not even capable of storing it all, despite the fact that storage is relatively inexpensive. Therefore, it is important to try and process the data on-the-fly as quickly as possible so that reduced sets of the original data can be stored, along with the results of the computation.” His hope is that now researchers can take advantage of newer machines as they become faster and more affordable, and smaller and more cost-effective.Genomics Started ItDouglas L. Brutlag, Professor Emeritus of Biochemistry and Medicine in the Department of Biochemistry at Stanford University, is currently involved in teaching computational molecular biology, bioinformatics, and genomic and proteomic courses. He says that growth in the field of bioinformatics and the increasing number of such departments at universities can be attributed to the field of genomics. Genomics, and the ability to predict the sequence of most proteins, will enable the systematic study of cell signaling, metabolism, and other pathways in whole organisms and individual cells simultaneously; and allow researchers to assign meaning not only to positive experiments, but to negative ones as well.Brutlag has identified computational goals of bioinformatics, which he thinks are approachable from either a biologic or engineering point of view. These goals include discovering and generalizing from probabilistic models of the sequences, structures, metabolism, and chemistries of well-studied systems, and to develop a systematic and genomic approach to molecular interactions, metabolism, cell signaling, and gene expression. Generalizations can be used for predictive models of gene expression, gene regulation, protein folding, protein-DNA interactions, protein-protein interactions, protein-ligand binding, catalytic functions, and metabolism. Ideally, models would take into account environmental conditions, whether they be those affecting an individual cell, the interaction of different tissues, or the organism in its milieu. “We don't have the time to do the laboratory work on all possible examples,” Brutlag observes, likening this process to machine learning in computer science. The engineering-like approach of synthetic biology is to attempt to construct novel organisms, functions, or regulation of genes and proteins in order to test predictive model systems. Another form of engineering involves creating mutations, inhibiting specific gene expression using RNAi, making genetic knock-outs, or using other means to target specific proteins to determine their function.Brutlag thinks studying predictive algorithms such as those used to simulate or predict protein folding, should lead to ways to simplify these algorithms and make them run faster. “Nature doesn't sit around and think about every atom, so we shouldn't either,” he says. However, those algorithms still must reflect the actual biology of a system. He's excited about the prospect of doing genome-wide association studies to correlate traits (such as diseases or susceptibilities) with single nucleotide polymorphisms (SNPs) and whole genes. He observes that although most of the “classical” diseases caused by a single genetic change (e.g., sickle cell anemia) have been identified, many diseases (e.g., diabetes, cancer, and heart disease) still require the study and correlation of many SNPs or genes. “I think we have to focus on disease genes, gene expression, and protein modification,” Brutlag says. “We have to understand these at the levels of DNA, RNA, and protein to understand the disease phenotype. We need to understand the effect of the environment, including the genetic environment, cell signaling pathways, and their component receptors and ligands, and work out all these pathways. The genome gives us the ability to do this.”FiguresReferencesRelatedDetailsCited ByUse of Omics Approaches for Developing Immune-Modulatory and Anti-Inflammatory Phytomedicines14 August 2015 Vol. 46, No. 7 Follow us on social media for the latest updates Metrics History Published online 25 April 2018 Published in print June 2009 Information© 2009 Author(s)PDF download