Current Bioinformatics - Volume 8, Issue 1, 2013

Protein-protein interactions play important roles in many biological progresses. Previous studies about proteinprotein interactions were mainly based on sequence analysis. As more 3D structural information can be obtained from protein-protein complexes, structural analysis becomes feasible and useful. In this study, we used structural alignment to predict protein-binding sites and analyzed interface properties using 3D alpha shape. We have developed a method for protein-protein interaction prediction. The result indicates good performance of our method in discriminating proteinbinding structures from non-protein-binding structures. In the experiment, our method shows best Matthews correlation coefficient of 0.204.

Cancer and diabetes are two highly malignant diseases. Accumulating evidence suggests that cancer incidence might be associated with diabetes mellitus, especially Type-2 diabetes which is characterized by hyperinsulinemia, hyperglycemia, obesity, and overexpression of multiple components of the WNT pathway. These diabetes risk factors can activate a number of signaling pathways that are important in the development of different cancers. To systematically understand the signaling components that link diabetes and cancer risk, we have constructed a single-cell, synchronous Boolean network model by integrating the signaling pathways that are influenced by these risk factors. Then, we have applied Model Checking, a formal verification approach, to qualitatively study several temporal logic properties of our diabetes-cancer model. Our aim was to study insulin resistance, cancer cell proliferation and apoptosis. The verification results show that the diabetes risk factors might not increase cancer risk in normal cells, but they will promote cell proliferation if the cell is in a precancerous or cancerous stage characterized by losses of the tumor-suppressor proteins ARF and INK4a.

Simultaneous activation of the agonist and antagonist muscles surrounding a joint, called co-contraction, is suggested to play a role in increasing joint stiffness to improve movement accuracy. However, it has not been clarified how co-contraction is related to movement accuracy, as most models for motor planning and control cannot deal with muscle co-contraction. In this study, the muscle activation and joint stiffness in reaching movements were studied under three different requirement levels of endpoint accuracy using a two-joint six-muscle model and an approximately optimal control. We carried out simulations of biological arm movements for a center-out reaching task under different accuracy demands with different types of motor noise and demonstrated time-varying co-contraction and a double-peaked jointstiffness profile. Furthermore, we showed that the strength of co-contraction and joint stiffness increased depending on the required accuracy level under signal-dependent noise, the magnitude of which was proportional to the motor command but not to additive Gaussian noise. We concluded that the optimal control is a valid model for the human motor control system and that signal-dependent noise is essential to induce co-contraction depending on accuracy demands.

This paper presents a novel speckle smoothing method for ultrasound images. This smoothing method is designed to preserve both the edges and structural details of the image. Speckle noise is suppressed by extending the smoothness of the image in the wavelet-based Hölder spaces. We try to solve the performance of speckle reduction problem from the viewpoint of Besov norm. A comparison of smoothing speckles with the other well-known methods is provided via the size of Besov norm. We validate the proposed method using synthetic data, simulated and real ultrasound images. Experiments demonstrate the performance improvement of the proposed method over other state-of-the-art methods in terms of image quality and edge preservation indices.

Gene regulation is a key factor in gaining a full understanding of molecular biology. microRNA (miRNA), a novel class of non-coding RNA, has recently been found to be one crucial class of post-transactional regulators, and play important roles in cancer. One essential step to understand the regulatory effect of miRNAs is the reliable prediction of their target mRNAs. Typically, the predictions are solely based on the sequence information, which unavoidably have high false detection rates. Recently, some novel approaches are developed to predict miRNA targets by integrating the typical algorithm with the paired expression profiles of miRNA and mRNA. Here we review and discuss these integrative approaches and propose a new algorithm called HCTarget. Applying HCtarget to the expression data in multiple myeloma, we predict target genes for ten specific miRNAs. The experimental verification and a loss of function study validate our predictions. Therefore, the integrative approach is a reliable and effective way to predict miRNA targets, and could improve our comprehensive understanding of gene regulation.

Most complex networks, such as biological networks, social networks, and information networks, are dynamic in nature. Therefore, analysis of these networks provides a better understanding of complex systems compared with analysis of static networks. In this paper, we define a new problem to find the substructures that are significant during the evolving periods, including conserved, appearing and disappearing substructures, and so on. We propose a novel framework for discovering such significant substructures. By using the representation of the summary graph and the approach of densitybased clustering algorithm, we just need to execute the core process in one network without the loss of the temporal information. Experiments on artificially generated PPI (protein-protein interaction) networks and real-world data show that our method can lead to the discovery of the significant substructures that reveal the dynamic local properties of dynamic networks. Also, our method can be used to analyze large scale networks.

Signaling pathways play central roles in responding to stimuli or transmitting signals from outside to inside of cells, which in turn regulate a series of complex biological processes that are vital for the function of cells. Unfortunately, the structure and function of most pathways are not complete or even not available. In the past decade, the availability of large amounts of high-throughput ‘omics’ data enables it possible to infer signaling pathways with computational methodologies that can help guide experiments in lab with low cost within short time. In this review, we present the latest progress being made in computational methodologies that are proposed for identifying signaling pathways from molecular interaction networks.

The expression of genes in three tissue types, adipose, liver and muscle, from normal (Wistar-Kyoto) and diabetic (Goto-Kakizaki) rats, was analyzed to gain insight into the potential functional relationships between these tissues in diabetes progression. For this purpose, the gene expression signatures during different periods of diabetes progression were first obtained, and were characterized by the biological functions. The significant functions for the three tissues were then clustered into 10 groups, according to the constituent expression profiles in the corresponding gene sets. Finally, the relationships between the 10 functional groups were inferred by a statistical network analysis method. The numbers of genes in the expression signatures were similar in the three tissues for the three periods, but the numbers of significant biological functions showed a wider variety for the three tissues. One of the remarkable features is that the functions of liver and muscle commonly detected in the three periods were found in those of adipose at the early stage. The following network analyses revealed a sketch of the functional relationships between the three tissues. The relationships within the three tissues were not similar, but those between them shared functional interfaces, in which adipose at an early stage and muscle during all periods transmit functional information, and liver at the late stage receives it, in diabetes progression. Thus, the present approach for functional network estimation may provide insights about the macroscopic relationships in diabetes progression, and especially the initiator role of adipose.

We designed a procedure to explore cellular relationships from sets of characteristic genes in distinctive cell types, by using a partial canonical correlation analysis. The present procedure was then applied to the characteristic gene sets of seven subtypes of testicular germ cell tumors, reported previously. The cellular relationships were reconstructed well, and were consistent with the general histologic lineage model. New implications for the subtype differentiation were found. In particular, the correspondence between the inferred relationships and the functional characterizations of constituent genes suggested a hypothesis for the classification between seminoma and embryonal carcinoma. The partial canonical correlation analysis is also appropriate for revealing new features of cellular relationships, based on the transcriptional programs. Thus, the present procedure facilitates the creation of a macroscopic view of the cellular relationships, by following the detection of characteristically expressed genes.

The induced pluripotent stem cells (iPSCs), generated from transcription factor-induced reprogramming, hold the great promise as the next generation materials for regenerative medicine. Intensive follow-up studies have accumulated a large amount of high-throughput data in transcription, proteomics, methylation, and other levels, which makes the computational studies feasible. Here we briefly review the recent bioinformatics efforts to study iPSCs. Specifically, we will summarize several comparison studies to determine how closely human iPSCs resemble human embryonic stem cells (ESCs) from sequence, gene expression profile, chromatin structure, DNA methylation, proteomics, and function aspects. Then computational methods to assess iPSC's pluripotency in a cost-effective yet accurate way are introduced. Finally, we will indicate the further biomolecular network studies to understand the underlying mechanism for cell reprogramming and the dynamics within this biological process.

Amyotrophic lateral sclerosis (ALS) is a complicated and devastating neurodegenerative disease. To date, its diagnosis is still mainly based on clinical symptoms and electromyographic findings. High rates of misdiagnosis and delayed diagnosis are the major hurdles in ALS treatment. Thus, searching for biomarkers to improve clinical diagnosis of ALS is a highly desirable goal. Here we review current potential biomarkers derived from the various pathogenic mechanisms of ALS, including those involved in oxidative stress, synaptic excitotoxicity, neuroinflammation and the autoimmune response. Oxidative stress results from genetic mutation or an increase in protein aggregation, synaptic excitotoxicity arising from elevated levels of glutamate and D-serine, and the neuroinflammation occurring from elevated levels of inflammatory molecules and cytokine receptors. Some of these biomarkers could be used for monitoring the disease progression and to assess effectiveness of treatment for ALS. We conclude that neuroinflammation plays a crucial role in ALS, which may lead to a better understanding of this devastating disease and ultimately to a cure. In addition, the identification of new biomarkers would undoubtedly provide critical insights into the pathogenesis of ALS.

Ion channels are integral membrane proteins that are responsible for most physiological functions including the electrical activity of excitable cells. Their expression and function are regulated by a variety of means including microRNAs (miRNAs). MicroRNAs are small non coding RNAs, 22 nucleotides in length, involved in regulation of gene expression. In this study, we attempt to predict the miRNA targets of ion channel genes conserved across 4 species using TargetScan algorithm. From the results, many miRNA targets were found to be conserved among mammals. The expression profile of identified miRNA targets of certain genes that are implicated in channelopathies of brain, heart and skeletal muscle in Homo sapiens are explored. Further a final correlation of channel genes and miRNAs expressed in specific tissues is obtained by comparison of expression profiles using mimiRNA. miR-302a and miR-9 were found to be positively correlated with many channelopathy associated genes, while miR-143 and miR-29a were predicted to exhibit negative correlation. In this paper we summarize a list of all possible miRNAs linked with channelopathies.

The histone octamer induced bending of DNA into the superhelix structure in nucleosome core particle is important for DNA packing into chromatin, which is directly associated with gene expression and transcription regulation. In this paper, we review current computational methodologies of nucleosome positioning signal recognition based on the sequence and structural features. Sequence-based methods can be effectively applied to genomic nucleosome mapping and genomic texture analysis around the nucleosome regions. The nucleosomal region has been proved to be more flexible than its surrounding regions. Structure-based methods incorporate physical and stereochemical properties of DNA molecules and employ signal processing and pattern recognition techniques to identify the structural significance hidden in the regularity of sequence composition. The special structural information of the specific dinucleotides in the nucleosome structure indicates their important roles in the formation of nucleosome complex. CA/TG shows the greatest flexibility in the structural variation along the DNA path, thus playing the role of “hinge” in the bending of DNA into superhelix. AA/TT and GC, on the other hand, have the highest probability of closely contacting with histone and play the role of facilitating nucleosome positioning.

Construction of the gene regulatory networks is a challenged problem in systems biology and bioinformatics. This paper presents construction of gene network using combined quantum-behaved PSO and K2 algorithm. Recent studies have shown that Bayesian Network is an effective way to learn the network structure. K2 algorithm is widely used because of its heuristic searching techniques and fast convergence, but it suffers from local optima. And the performance of K2 algorithm is greatly affected by a prior ordering of input nodes. Quantum-behaved PSO is a population-based stochastic search process, which automatically searches for the optimal solution in the search space. So, we combined it with K2 algorithm for construction gene network. The results of hybrid PSO, K2 (we refer to it as QPSO-K2 algorithm), stand-alone K2 and quantum-behaved PSO algorithms are compared on several datasets. Among the three algorithms, the hybrid QPSO-K2 algorithm performs well for all of the datasets.