Current Bioinformatics - Volume 5, Issue 3, 2010

Epigenetics has become a cornerstone of cancer research and is an increasingly important factor in the continuous efforts to try and unravel the biology of oncogenesis. Consequently the analyses of epigenetic data have evolved towards genome wide and high-throughput approaches, generating large data sets for which computational data mining is indispensable. Bioinformatics has proven to be useful and beneficial for a plethora of tasks, going beyond elemental data management, and is now crucial for adequate candidate gene selection, data integration, comparison and correlation as well as providing insights into cancer biology. Computational approaches are used even in routine tasks like primer design, since multiple layers of information can be incorporated into a more efficient and consistent strategy. Almost every analysis feeds back information into both the biology and the tools we use during these experiments. As the cancer epigenetics field evolves rapidly, the combination with bioinformatics will create a synergy that increases our insights into cancer biology rapidly. This review summarizes some of the frequently used bioinformatics tools in large-scale or nextgeneration analyses in epigenetics that would not have been possible without the use of well-conceived computational strategies.

The goal of tumor immunology is to understand the interactions between tumor cells and immune system, and ultimately to devise immune basedapproaches to fight cancer. We discuss here recent advances in tumor immunology and in the interaction between immunology and informatics that provide new perspectives for the development of cancer therapy. Cancer immunoprevention is a novel approach to cancer prevention through the use of vaccines and other immunological strategies to be applied before tumor onset. The efficacy of cancer immunoprevention has been demonstrated in several experimental systems, however, as frequently happens in novel approaches, cancer immunoprevention studies incorporate a large number of variables. We show here how a specifically designed lattice gas model can provide significant insight for the analysis of immune variables and the design of new biological experiments. A second approach that provides new perspectives in tumor immunology as the result of interactions between immunology and informatics is the use of DNA microarrays to investigate and monitor tumor-host relationships and modifications induced by immunopreventive and immunotherapeutic interventions.

There has been substantial evidence that functional and structural analyses of genes and proteins can help develop new drugs, diagnose medical conditions and find cures for diseases. However, these biological sequence analyses require large-scale computational power due to the exponential growth of genomic information. During the past two decades considerable efforts have been expended in trying to accelerate the biological sequence database search process which is the fundamental step for further analyses. Various software approaches including SIMD (Single Instruction Multiple Data) instruction, multithreading, message passing programming paradigm and I/O optimization have been employed to speed up the process on different computing platforms at different levels. Hardware techniques such as FPGA (Field Programmable Gate Arrays), GPU (Graphics Processing Unit), IBM CELL BE, DSP (Digital Signal Processors) and ASIC (Applications Specific Integrated Circuit) have also been widely used. This paper reviews relevant computing platforms, various software and hardware approaches as well as the performances they achieved in high throughput sequence database search. It demonstrates that parallelism can be exploited at different phases, granularity levels, types, software/ hardware levels and scopes. This would help researchers understand current development strategies and possible future trends such as aggregate heterogeneous systems in high performance biological sequence analysis.

Predicting the thermostability of a biomolecule, given its sequence, is one of the big challenges of protein engineering and developing tools to screen thermostable mutants is of great interest. Here we used various screening, clustering, decision tree and generalized rule induction models to search for patterns of thermostability. Arg was solely found as N-terminal amino acid in proteins at temperatures higher than 70°C. Fifty-four protein features were important in feature selection, and the number of peer groups (anomaly index 2.12) declined from 7 to 2 with selected features; no changes were found in K-Means and TwoStep clusters with/without feature selection filtering. Tree depths of decision tree models varied from 14 (in C5.0 with 10-fold cross-validation and with feature selection) to 4 (in CHAID) branches and C5.0 was the best and the Quest model was the worst. No significant difference in the performance of various decision tree models was found with/without feature selection, but the number of peer groups in clustering models was reduced significantly (p<0.05). The frequency of Gln was the most important feature in decision tree rules and for all association rules in antecedent to support the rules. The importance of Gln in protein thermostability is discussed in this paper.

Microarray technology enables the monitoring of the expression patterns of a huge number of genes across different experimental conditions or time points simultaneously. Biclustering of microarray data is an important technique to discover a group of genes that are co-regulated in a subset of experimental conditions. Traditional clustering algorithms find groups of genes/conditions over the complete feature space. Therefore they may fail to discover the local patterns where a subset of genes has similar behaviour over a subset of conditions. Biclustering algorithms aim to discover such local patterns from the gene expression matrix, thus can be thought as simultaneous clustering of genes and conditions. In recent years, a large number of biclustering algorithms have been proposed in literature. In this article, a study has been made on various issues regarding the biclustering problem along with a comprehensive survey on available biclustering algorithms. Moreover, a survey on freely available biclustering software is also made.

The theoretical prediction of protein structures has become a field of increasing importance in both biology and physics. Reliable prediction methods in fact, would spare time consuming experimental X-ray and NMR techniques and they would represent a challenge for computational protein modeling as well. The well known limitations of all-atom models call for the development of coarse-grained protein descriptions including a minimal number of protein-like features, while being capable of mimicking the essence of protein folding mechanisms. In this paper we review the most important classes of coarse-grained protein models in order of increasing complexity, starting from (over simplified) binary models, to models with one or two reaction centers per residue. We discuss how, despite their simplification, coarse-grained models constitute a viable approach to structure prediction and they shed light on many aspects of protein-folding problem.