Current Bioinformatics - Volume 1, Issue 1, 2006

The drug discovery process is complex, time consuming and expensive, and includes preclinical and clinical phases. The pharmaceutical industry is moving from a symptomatic relief focus towards a more pathology-based approach where a better understanding of the pathophysiology should help deliver drugs whose targets are involved in the causative processes underlying the disease. Computational biology and bioinformatics have the potential not only to speed up the drug discovery process, thus reducing the costs, but also to change the way drugs are designed. In this review we focus on the different computational and bioinformatics approaches that have been proposed and applied to the different steps involved in the drug development process. The development of 'network-reconstruction' methods is now making it possible to infer a detailed map of the regulatory circuit among genes, proteins and metabolites. It is likely that the development of these technologies will radically change, in the next decades, the drug discovery process, as we know it today.

Current projects for the massive characterization of proteomes are generating protein sequences and, to less extent, three dimensional structures with unknown function. Experimentally determining functional features of a protein is expensive, time consuming and difficult to automate. There is therefore a demand for computational methods for predicting protein functional features, which can be coupled to the pipelines of genome sequencing and structure determination. This review focuses on current in-silico methods for predicting regions in proteins with some functional importance (catalytic sites, binding sites, protein interaction regions, etc.) using sequence and/or three-dimensional structure information.

The interest in finding the keys to the thermal stabilization of proteins has remained constant and unquestionable throughout the last twenty years. This article reviews the most recent theoretical and computer advances related to the problem of thermally stabilizing proteins. Although comparison between mesophilic and thermophilic sequences has suggested some thermostabilization mechanisms, it has not been able 'per se' to provide unambiguous thermostabilization rules applicable for every case. Two of the mechanisms used by nature are seen as the major factors governing thermostability: the electrostatic forces of charged amino acids within a protein and the packing of its hydrophobic core on the other. Other mechanisms that have also been implicated (i.e. hydrogen bonding, α-helix stabilization, backbone rigidifying, etc), may play a refining role, based on the principle that nature has punctually and opportunistically thermostabilized proteins in each particular case, thereby solving each specific problem. How electrostatic and hydrophobic forces affect each other is still remains a largely open question and some recently developed criteria based on these two effects have been analyzed in the review.

Proteome analysis linked to genome sequence information is very useful for functional genomics. Since proteins are the major players in most processes of living cells, knowledge of the proteome has great relevance to the study of cells and organisms at the molecular level. The technique of proteome analysis using two-dimensional polyacrylamide gel electrophoresis (2D-PAGE) has the power to monitor global changes that occur in the protein complement of tissues and subcellular compartments. As a complement to more focused studies, and to facilitate further advances in functional genomics, several databases based on 2D-PAGE are already available including those for plants. In this review, the rice proteome database and other plant proteome databases are discussed. Organizing and streamlining the access of information into plant proteome databases, especially the rice proteome database, will aid in cloning the genes and predicting the function of unknown proteins.

Microarrays provide the biological research community with tremendously rich, sensitive and detailed information on gene expression profiles. Gene expression profiling and gene expression patterns have been found useful for solving a wide variety of important biological and biomedical problems, including the study of metabolic pathways, inference of the functions of unknown genes, diagnosis of diseased states, as well as facilitating the development of individualized drug treatments through pharmacogenomics. Given the significant impact of microarray gene expression data in biological and biomedical research, this breakthrough technology urgently needs the assistance of advanced computational methods for interpreting and utilizing the raw information. This paper reviews several main research directions and methods in the analysis of microarray gene expression data.

In this review, we have discussed the class-prediction and discovery methods that are applied to gene expression data, along with the implications of the findings. We attempted to present a unified approach that considers both class-prediction and class-discovery. We devoted a substantial part of this review to an overview of pattern classification/recognition methods and discussed important issues such as preprocessing of gene expression data, curse of dimensionality, feature extraction/selection, and measuring or estimating classifier performance. We discussed and summarized important properties such as generalizability (sensitivity to overtraining), built-in feature selection, ability to report prediction strength, and transparency (ease of understanding of the operation) of different class-predictor design approaches to provide a quick and concise reference. We have also covered the topic of biclustering, which is an emerging clustering method that processes the entries of the gene expression data matrix in both gene and sample directions simultaneously, in detail.

Databases of three-dimensional macromolecular structures became so large that fast search tools and comparison methods were needed and were actually designed. All of them employ simplified representations of the threedimensional structure: strings of characters of variable length, which can be handled with procedures that were designed for sequence analysis; fixed dimension arrays that can be processed with standard statistical methods; ensembles of secondary structural elements, which are much less numerous than the atoms/residues of the protein; and continuous representations of the backbone, through stereochemical figures. Some of these computational procedures were developed long ago, when computers were too slow, and others have been designed recently, with the specific aim of handling large amount of information. The present article is focused on the algorithms that allow fast structure comparison, particularly suitable to handle large databases, and should provide a comprehensive picture, useful for the development and the assessment of novel tools.

Our understanding of biological systems has improved dramatically due to decades of exploration. This process has been accelerated even further during the past ten years, mainly due to the genome projects, new technologies such as microarray, and developments in proteomics. These advances have generated huge amounts of data describing biological systems from different aspects. Still, integrating this knowledge to reconstruct a biological system in silico has been a significant challenge for biologists, computer scientists, engineers, mathematicians and statisticians. Engineering approaches toward integrating biological information can provide many advantages and capture both the static and dynamic information of a biological system. Methodologies, documentation and project management from the engineering field can be applied. This paper discusses the process, knowledge representation and project management involved in engineering approaches used for biological information integration, mainly using software engineering as an example. Developing efficient courses to educate students to meet the demands of this interdisciplinary approach will also be discussed.

Recent progress in experimental techniques such as high-throughput genome sequencing, proteomics, transcriptomics and interactomics have lead to a new demand for integrated computational analyses, capable of systematically organizing these heterogeneous, fragmentary data into a coherent whole. As a consequence, novel systemlevel bioinformatics solutions are now being developed with the goal of understanding and predicting the behaviour of complex systems, such as molecular pathways, cells, tissues, organs and even whole organisms. Multiple alignments of both nucleotide and protein sequences play a central role in many of these applications, which range from the identification of genes and their products, via the characterisation of their 3D structure and their molecular and cellular functions, to the prediction of the phenotypic consequences of mutations, reverse engineering and drug design. In a multiple sequence alignment, structural and functional data can be combined with evolutionary information to allow reliable data validation, consensus predictions and rational propagation of information from known to unknown sequences. Clearly, integration at this scale calls for high quality, automatic multiple alignments. Alignment techniques are now responding to the challenge, with current developments moving away from a single all-encompassing algorithm towards more co-operative, knowledge based systems. However, the success of these methods relies on the efficient integration of information from different databases and the close cooperation of the different data mining and investigation algorithms. A large community effort is now underway to develop standards for data exchange and organisation that will facilitate collaborations between the various resources, in order to support improved domain understanding and to provide better decision-making systems and services for the biologist.

One of the main topics in genomics is to determine the relevance of DNA variations with some genetic disease. Single nucleotide polymorphism (SNP) is the most frequent and important form of genetic variation which involves a single DNA base. The values of a set of SNPs on a particular chromosome copy define a haplotype. Because of its importance in the studies of complex disease association, haplotyping is one of the central problems in bioinformatics. There are two classes of in silico haplotyping problems, i.e., single individual haplotyping and population haplotyping. In this review paper, we give an overview on the existing models and algorithms on this topic, report the recent progresses from the computational viewpoint and further discuss the future research trends.