Computational Biology: Challenges and Opportunities
The current issue of the quarterly publication, CTWatch, focuses on the issues and challenges facing the field of computational biology today and in the future. A recurring theme throughout all of the articles is that the field of biology is becoming increasingly data driven and is producing data faster than computers can process it. The authors address the limitations of our current cyberinfrastructure and suggest strategies to overcome these challenges.
In his introduction, “Trends in Cyberinfrastructure for Bioinformatics and Computational Biology,” Rick Stevens, Associate Laboratory Director, Computing and Life Sciences of Argonne National Laboratory and Professor, Computer Science Department of The University of Chicago, outlines three major trends in biology research: the increasing availability of high-throughput data, the acceleration of the pace of questions whose answers rely on increasing computation resources, and simulation and modeling technologies that will eventually lead to predictive biological theory.
Stevens addresses the role of petascale computing with regard to fundamental biological problems, such as the evolutionary history of genes and genomes. This is significant, as the number of completed genome sequences will reach 1,000 in the next few years. He provides a list of multiple “problem areas” and their estimated time to completion at three levels of computing power (360, 1000, and 5000 teraflops). For example, on the IBM Blue Gene/L, screening “all known microbial drug targets against the public and private databases of chemical compounds to identify potential new inhibitors and potential drugs,” would take one year for all microbial targets at 360 teraflops, a one month for all microbial targets at 1000 teraflops, and one machine year for all known human drug targets at 5000 teraflops.
Eric Jakobsson of the National Center for Supercomputing Applications, University of Illinois at Urbana-Champaign discusses the initiatives that will be required between applications scientists and system architects in order to come up with a suitable cyberinfrastructure for biology in “Specifications for the Next-Generation Computational Biology Infrastructure.” One of the five integration models Jakobsson outlines is “Integration of algorithmic development with computing architecture design.” He says:
“The different types of biological computing have vastly different patterns of computer utilization. Some applications are very CPU-intensive, some require large amounts of memory, some must access enormous data stores, some are much more readily parallelizable than others, and there are highly varied requirements for bandwidth between hard drive, memory, and processor.”
Jakobsson suggests that more extensive mutual tuning of computer architecture to applications software would make existing and projected computational resources more productive. One case of such tuning is the molecular simulation code Blue Matter, designed to leverage the architecture of the IBM Blue Gene supercomputer. Jakobsson praises the Blue Matter-Blue Gene combination, declaring that it has enabled important new discoveries.
Jakobsson also calls for better training in the area of computational biology at the undergraduate and graduate levels. He points to the University of California at Merced as one institution that has fully integrated computing into all levels of its biology curriculum as called for in the National Academy of Sciences BIO 2010 report.
In “Genome Sequencing vs. Moore's Law: Cyber Challenges for the Next Decade” Folker Meyer of the Argonne National Laboratory addresses the challenge of the number of sequenced genomes growing faster than Moore's Law. He states that the number of available complete genomic sequences is doubling every 12 months, faster than Moore's 18 months. “The analysis of genomic sequences requires serious computational effort: most analysis techniques require binary comparison of genomes or the genes within genomes. Since the number of binary comparisons grows as the square of the number of sequences involved, the computational overhead of the sequence comparisons alone will become staggering.”
As the number of sequences grows so do the number of algorithms to study them, requiring additional computer power. For example, using Hidden Markov Models to search for sequence similarities not visible with the traditionally used BLAST algorithm requires greater computing resources. The author states that the TeraGrid is one of the few resources that can handle the computational requirement. We need to overcome these limitations in order to study and better understand “crop plants, pathogens and ultimately human beings.” To resolve the gap between data and resource, the author calls for new bioinformatics techniques as well as high-throughput computing, concluding that “biology is in the middle of a paradigm shift towards becoming a fully data driven science.”
In “Computing and the “Age of Biology,' ” Natalia Maltsev of the Argonne National Laboratory calls for the “development of high-throughput computational environments that integrate (i) large amounts of genomic and experimental data, (ii) comprehensive tools and algorithms for knowledge discovery and data mining, and (iii) comprehensive user interfaces that provide tools for easy access, navigation, visualization, and annotation of biological information.” For achieving this integrated environment, Maltev makes four recommendations.
First, she calls for large, public, scalable computational resources to handle the exponential growth of biological data. For example, the largest genomic database, GenBank, contains 56 billion bases, from 52 million sequences; and as the cost of sequencing new genomes drops, the rate of growth of GenBank is expected to increase dramatically.
Second, Maltev proposes a new model to handle the increasing complexity of biological data. She states that biology is becoming increasingly multi-disciplinary, “using information from different branches of life sciences; genomics, physiology, biochemistry, biophysics, proteomics, and many more.” The model needs to incorporate various classes of biological information as well as similar classes of data from different resources. According to Maltev, the difficulty with an integrated model is due to “the large volume and complexity of data, the distributed character of this information residing in different databases, shortfalls of current biological ontologies, and generally poor naming conventions for biological objects.”
Maltsev's third recommendation is algorithm development. The current bioinformatic tools (for example, BLAST and FASTA) are not adequate to handle the exponential growth of sequence data. Maltev says “bioinformatics will significantly benefit from the development of a new generation of algorithms that will allow efficient data mining and identification of complex multidimensional patterns involving various classes of data.”
Maltev's fourth and final recommendation is the development of collaborative environments that will allow researches in different locations to view and analyze the data. Maltev claims that storing data and its analysis in one location will not meet the needs of biology in the future. She also calls for visualization of information to reduce its complexity.
Maltev's article provides an accessible framework for understanding the challenges of computational biology. In the “age of biology,” computing and biology will unite to solve major global problems such as curing deadly diseases and ending world hunger.
The message in all of these articles is that biology has become a data-driven discipline and is becoming increasingly more so. Computational resources cannot keep up with the data, and questions are piling up faster than answers. Remedying this situation is essential for progress.
To view the complete issue of CTWatch, visit their website at http://www.ctwatch.org/.