Applied Bioinformatics Computing: An Introduction

Like this article? We recommend

Bioinformatics Computing

Learn More Buy

Like this article? We recommend

Like this article? We recommend 

Bioinformatics Computing

Learn More Buy

The Central Dogma

The Central Dogma of Molecular Biology was originally defined by the American biochemist James Watson who, together with the British physicist Francis Crick, first described the now famous right-handed double helix of DNA (deoxyribonucleic acid) in 1953. The Central Dogma is deceptively simple: DNA defines the synthesis of protein by way of an RNA intermediary. What isn't so simple is documenting, controlling, and modifying this process (illustrated in Figure 2), which is the focus of much of bioinformatics.

Figure 2 The Central Dogma: DNA is transcribed to RNA, which is translated to protein.

As shown in Figure 2, the replication of DNA and the transcription of DNA to RNA occurs in the cell nucleus, which houses the DNA in the form of tightly-packed chromosomes. The translation of RNA to protein (the building block of everything from blood to the muscles and organs of the body) occurs in the cytoplasm.

From a computer science perspective, the biology of this process may be less important than the flow of data it represents. For example, the process is inherently digital with a four-character alphabet (A, T, C, and G)—with each character representing a nucleotide or base. Following the message from the original DNA in the nucleus to protein in the cytoplasm, A combines with T, and C combines with G—at least in theory. In practice, however, the A-T and C-G base pairings aren't always exact. There is an error rate of about one base pair mistake in every million base pairs. Like a single bit error in a computer program, the effect of the error might be insignificant or horrific, depending on exactly where the error occurs. A single error may result in debilitating diseases such as sickle cell anemia or cystic fibrosis, for example.

Understanding these and other errors so that they can be corrected is one focus of bioinformatics. It's important to note that in attaining this understanding, modeling the process according to Claude Shannon's theory of communications is necessary but insufficient. Although Shannon's theory aptly specifies the amount of information that can be transferred from DNA in the nucleus to the protein synthesis machinery in the cytoplasm as a function of the noise level of the cellular matrix, it ignores the biology of the system. For example, people who carry one copy of the defective sequence definition that results in sickle cell anemia are relatively resistant to malaria. As a result, people who live in areas of the world in which mosquitoes carry the parasite that results in malaria benefit from what we consider a disease in the malaria-free U.S. The point is that although many biological systems can be reduced to relatively simple and mathematically sound models, knowledge of the relevant biology is needed to fully appreciate the applicability of particular computer science methods.

Databases

Bioinformatics is characterized by an abundance of data stored in very large databases. Local databases with capacities measured in the tens of terabytes are common. As such, fluency in data warehousing, data dictionaries, database design, archiving, and knowledge management techniques are mandatory for the design and maintenance of these systems. Most of the large biology databases are based on traditional relational databases architectures; whereas others, especially systems dealing with images and other multimedia, are based on object-oriented designs.

A sample of the types of databases available online is listed in Table 1. Readers who aren't familiar with database types are encouraged to read the ancillary materials that accompany many of the systems. The tutorial materials that accompany the larger systems, such as the biomedical literature database PubMed (http://www.ncbi.nlm.nih.gov/entrez/query.fcgi) or the image-rich Protein Data Bank or PDB (http://www.rcsb.org/pdb/), are particularly informative.

Table 1 Popular public online databases serving the bioinformatics community.

In addition to database architecture and design, computer science professionals in charge of the online biology databases, as well as those who are charged with developing and maintaining local data warehouses, must be conversant in database management, data lifecycle management, computer interface development, and implementation of large database projects on time and on budget. In this regard, the database component of bioinformatics differs little from database projects in the banking, retail, or medical fields.

Networks

Bioinformatics, like virtually every other knowledge-intensive field, is dependent on a robust information technology infrastructure that includes the Internet, the World Wide Web, intranets, and wireless systems. These and other network technologies are applied directly to sharing, manipulating, and archiving genetic sequences and other bioinformatics data. For example, the majority of resources available for researchers in the bioinformatics are Web-based systems such as GenBank, which is maintained by the National Center for Biological Information (NCBI), the National Institutes of Health (NIH), and other government agencies.

The issues and challenges associated with providing an adequate network infrastructure are related to selecting and implementing the appropriate communications models, selecting the best transmission technology, identifying the most effective protocols, dealing with limited bandwidth, selecting the most appropriate network topologies, and contending with security and privacy. Because of the computational requirements associated with bioinformatics, the field serves as a test-bed for many of the leading-edge networking technologies, such as the Great Global Grid (GGG), which distributes not only data but supercomputing-level processing power to PCs and workstations as well.

Search Engines

The exponentially increasing volume of data accessible in digital form over the Internet—from gene sequences to published references in the biomedical literature to the experimental methods used to determine specific sequences—is accessible only through advanced search engine technologies. In this regard, the challenges faced by designers of bioinformatics-specific search engines are virtually identical to those addressed by computer scientists working in other areas. These include how to best constrain the search space, how to use hashing and other pre-processing methods to increase performance, and how to combine powerful search engine technologies in a manner that is not only powerful but also usable.

Visualization

Most molecular biologists agree that protein function is related to form. Unfortunately, experimental methods of determining protein structure, such as X-Ray crystallography and Nuclear Magnetic Resonance (NMR) imaging, are typically painstakingly slow and expensive—hence the interest in predicting protein structure through computational methods.

Exploring the possible configurations of folded proteins has proven to be virtually impossible by simply studying linear sequences of bases. However, sophisticated 3D visualization techniques allow researchers to use their visual and spatial reasoning abilities to understand the probable function of proteins. For example, the molecule featured in Figure 3: a form of human insulin. The structure is derived from data in the Protein Data Bank (PDB) that is rendered with freely available software that can be run within a Web browser or downloaded to take advantage of local processing power. By using tools that allow the protein to be rotated in virtual free space, scientists can experiment with the interaction of protein molecules and identify potential interactions in lieu of using arduous experimental wet-lab methods.

Figure 3 Rendering of human insulin. From the PDB (Protein Data Bank) Structure Explorer, based on MolScript and Raster3D. The two superimposed spheres in the center of the figure represent zinc ions.

Statistics

The randomness inherent in any sampling process, including measuring the reactions of thousands of genes simultaneously with microarray techniques or assessing the similarity between genetic sequences, necessarily involves probability and statistical methods. In many cases, the statistical techniques applied to bioinformatics problems are integrated within other applications and support activities such as the statistical analysis of structural features, gene prediction, and quantifying uncertainty in sequencing results.

Data Mining

In the early 1980s, gene sequencing worldwide resulted in about four base pairs per day. Today, scientists worldwide are contributing about 1,000 base pairs per second to the online sequence databases. Given this ever-increasing store of sequence and protein data from several ongoing genome projects, data mining the sequences has become a field of research in its own right. Thanks to the ongoing development of data mining applications, many scientists are able to conduct significant basic research from their Web-connected PC, without ever entering a wet lab or seeing a sequencing machine.

In addition to mining the sequence databases, many researchers are developing powerful text-mining applications that are capable of extracting data from online biomedical literature databases such as PubMed. Many areas are still out of reach for these and other traditional text-mining methods. For example, although algorithms are available to summarize a multi-page document into a single paragraph that can be quickly reviewed, the content contained in images and tables in the document are lost in the summarization process.

Pattern Matching

Classical pattern matching through standard AI techniques—such as reasoning under uncertainty, machine learning, image and pattern recognition, neural networks, and rule-based expert systems—have direct and practical applicability to practical bioinformatics research and development. For example, real-time microarray analysis lends itself to machine learning, in that it is humanly impossible to follow tens of thousands of parallel reactions unaided. Similarly, several gene prediction applications in bioinformatics are based on neural network pattern-matching engines.

One of the most often-used pattern-matching approaches in bioinformatics is dynamic programming, which is essentially recursive programming with a memory of intermediate results. Dynamic programming is used to align sequences that don't exactly match, but that are close enough to suggest that the two molecules considered in the alignment are similar in form(and, by extension, perhaps function). In other words, two molecules of the same general shape and configuration may be related evolutionarily (homology).

However, even if they aren't related, they likely behave similarly in the body because they share a structure. For example, the structure of the hemoglobin molecule found in a monkey blood is virtually identical to the hemoglobin molecule found in human blood, and both perform essentially the same function in each system.

Modeling and Simulation

Modeling and simulation are essential for understanding any complex system, and exploring the inner workings of a cell at the molecular level is no exception. A variety of simulation techniques is used in bioinformatics to model potential drug-protein interactions, probable protein folding configurations, and the analysis of potential biological pathways. Modeling and simulation techniques are most useful when they are linked with visualization techniques.

Collaboration

Bioinformatics is characterized by a high degree of cooperation between the researchers who contribute their part to the whole knowledge base of genomics and proteomics. This level of collaboration is made possible by technologies that facilitate multimedia communications, such as real-time videoconferencing, groupware, Web portals for submission of sequence data, and the Internet, of course.