Current Bioinformatics - Volume 1, Issue 3, 2006

In a gene network, genes may be expressed constantly, or expressed based on molecular signals. Transcription is a key process in gene expression. Through evolution, biological organisms have developed internal regulatory mechanisms for transcription control. Such mechanisms dictate how the network will function under certain environmental conditions and respond to changes in the environment. To develop formal approaches that enable the design and synthesis of such logical controls in artificial gene networks represents a major challenge. A first step in meeting this challenge would be to build analytical models of transcription control. This paper reviews computational approaches for modeling transcription control in gene networks from a systems-theoretic perspective, with emphasis on the logical representational capability of the models and their potential use in synthesis of external control.

In recent years, the broad utilization of high-throughput experimental techniques resulted in a vast amount of expression and interaction data, accompanied by information on metabolic, cell signaling and gene regulatory pathways accumulated in the literature and databases. Thus, one of the major goals of modern bioinformatics is to process and integrate heterogeneous biological data to provide an insight into the inner workings of a cell governed by complex interaction networks. The paper reviews the current development of semantic network (SN) technologies and their applications to the integration of genomic and proteomic data. We also elaborate on our own work that applies a semantic network approach to modeling complex cell signaling pathways and simulating the cause-effect of molecular interactions in human macrophages. The review is concluded with a discussion of the prospective use of semantic networks in bioinformatics practice as an efficient and general language for data integration, knowledge representation and inference.

A major breakthrough in classifying proteins from different microbial genomes in terms of sequence similarity was the development of the COG concept by Tatusov et al. in 1997. The authors defined clusters of orthologous groups of proteins (COGs) by strictly applying all against all BLAST alignments of protein sequences from completely sequenced microbial genomes. The latest update of the COG database already covered 66 microbial genomes and additionally included the KOG database, an equivalent consisting of seven eukaryotic genomes. Although excellent web-based software tools designed to analyze this huge amount of data were initially provided by the authors, many other groups independently developed more specialized or extended programs making use of COG data for diverse purposes. Here a brief introduction is given to the concept behind COGs and their potentials in the field of comparative and functional genomics are discussed. The review then is focused on the multitude of recently developed web services aimed at mining the COG database. Their capabilities to solve diverse problems in biochemistry are addressed. In order to illustrate the broad field of possible applications, a compilation of recently published findings, implementing information derived from comparative genomics with emphasis on data retrieved from the COG database, is given.

Inferring comprehensive regulatory networks from high-throughput data is one of the foremost challenges of modern computational biology. As high-throughput expression profiling experiments have gained common ground in many laboratories, different techniques have been proposed to infer transcriptional regulatory networks from them. Furthermore, with the advent of diverse types of high-throughput data, the research in network inference has received a new impulse. The use of diverse types of data, together with the increasing tendency of building the inference on biologically plausible simplifications, allows a more reliable and more complete description of networks. Here, we discuss how the research focus in the field of network inference is increasingly shifting from methods trying to reconstruct networks from a single data type towards integrative approaches dealing with several data sources simultaneously to infer regulatory modules.

Computational modeling and simulation have become invaluable tools for the biological sciences. Both aid in the formulation of new hypothesis and supplement traditional experimental research. Many different types of models using various mathematical formalisms can be created to represent any given biological system. Here we review a class of modeling techniques based on particle-based stochastic approaches. In these models, every reacting molecule is represented individually. Reactions between molecules occur in a probabilistic manner. Modeling problems caused by spatial heterogeneity and combinatorial complexity, features common to biochemical and cellular systems, are best addressed using Monte-Carlo single-particle methods. Several software tools implementing single-particle based modeling techniques are introduced and their various advantages and pitfalls discussed.

Discovery of transcription regulatory elements has been an enormous challenge, both to biologists and computational scientists. Over the last three decades, significant progress has been achieved by various laboratories around the world. Earlier, laborious experimental methods were used to detect one or handful of elements at a time. With recent advances in DNA sequencing technology, many completed genomes became available. High throughput biological techniques and computational methods emerged. Comparative genomic approaches and their integration with microarray gene expression data provided promising results. In this review, we discuss the development of technology to decipher the complex transcription regulation system with a focus on the discovery of cis-regulatory elements in eukaryotes.

We have conducted a dedicated analysis on the frequency distribution of the TATA Box and TATA extension sequences on six data sets of human promoters. Promoters in these sets have different lengths and are from different types of genes (housekeeping genes, tissue specific genes, and all genes). The statistical approach developed in this study will firstly partition the promoters into bins of 20 bp long, then calculate the frequency distribution of TATA elements and TATA extension sequences. The median value is used to capture outstanding TATA elements or TATA extension sequences when calculating their statistical significance. This study discovered that two of the 16 TATA Box elements (TATAAAAG and TATATAAG) showed the sharpest peaks at the location of 10∼30 bp upstream from transcription start sites where TATA Box is believed to reside. Fourteen TATA Box extensions showed the sharpest peaks at this location as well among all TATA extension sequences. Two of these fourteen TATA extension sequences have been verified to be the transcription factor binding sites by other research efforts. We suggest that the remaining twelve TATA extension sequences are the new putative TATA binding sites. This study also found that there was very little difference between the frequency distribution of TATA elements on housekeeping genes and their frequency distribution on tissue specific genes.

To a great extent, our phenotype is determined by our genetic material. Many genotypic modifications may ultimately become manifest in more or less pronounced changes in phenotype. Despite the importance of how specific genetic alterations contribute to the development of diseases, surprisingly little effort has been made towards exploiting systematically the current knowledge of genotype-phenotype relationships. In the past, genes were characterized with the help of so-called "forward genetics" studies in model organisms, relating a given phenotype to a genetic modification. Analogous studies in higher organisms were hampered by the lack of suitable high-throughput genetic methods. This situation has now changed with the advent of new screening methods, especially RNA interference (RNAi) which allows to specifically silence gene by gene and to observe the phenotypic outcome. This ongoing large-scale characterization of genes in mammalian in-vitro model systems will increase phenotypic information exponentially in the very near future. But will our knowledge grow equally fast? As in other scientific areas, data integration is a key problem. It is thus still a major bioinformatics challenge to interpret the results of large-scale functional screens, even more so if sets of heterogeneous data are to be combined. It is now time to develop strategies to structure and use these data in order to transform the wealth of information into knowledge and, eventually, into novel therapeutic approaches. In light of these developments, we thoroughly surveyed the available phenotype resources and reviewed different approaches to analyzing their content. We discuss hurdles yet to be overcome, i.e. the lack of data integration, the missing adequate phenotype ontologies and the shortage of appropriate analytical tools. This review aims to assist researchers keen to understand and make effective use of these highly valuable data.

The genome projects in human and other species have made genetic data widely available and pose challenges as well as opportunities for statistical analysis. In this paper we elaborate the concept of integrated analysis of genetic data, such that most aspects of analyses can be done effectively and efficiently in environments with facility for database accessibility, graphics, mathematical/statistical routines, flexible programming language, re-use of available codes, Internet connectivity and availability. This extends an earlier discussion on software consolidation (Guo and Lange. Theor Pop Biol 57:1-11, 2000). A general context is laid out by recollecting the research paradigms for genetic mapping of complex traits and illustrated with the study of ageing, before turning to the computational tools currently used. We show that the R system (http://www.r-project.org) so far is the most comprehensive and widely available system. However, other commercial systems can potentially be successful. In particular, we compare SAS (http://www.sas.com), Stata (http://www.stata.com), S-PLUS (http://www.insightful.com) and give some indications of future development. Our investigation has important implications for both statisticians end other researchers actively engaged in analysis of genetic data.