Dysregulation of epigenetic mechanisms has been implicated in the pathogenesis of Alzheimer's disease (AD). It has been shown that epigenetic status in promoter regions can alter levels of gene expressions, but their influence on correlated expressions of genes and its dependency on the disease are unclear. Using publicly available microarray and DNA methylation data, this article infer how correlation in gene expression in non-demented (ND) and AD brain may be influenced by genomic promoter methylation. Pearson correlation coefficients of gene expression levels between each of 123 known hypomethylated genes and all other genes in the microarray dataset were calculated, and the mean absolute coefficients were obtained as an overall strength of gene expression correlation of the hypomethylated gene. The distribution of the mean absolute coefficients showed that the hypomethylated genes can be divided into two, by the mean coefficients above or below 0.15. The division of the hypomethylated genes by the mean coefficients was more evident in AD brain than in ND brain. On the other hand, hypermethylated genes had a single dominant group, and the majority of them had the mean coefficient below 0.15. These results suggest that the lower the DNA methylation, the higher the correlation of gene expression levels with the other genes in microarray data. The strength of gene expression correlation was also calculated between known AD risk genes and all other genes in microarray data. It was found that AD risk genes were more likely to have the mean absolute correlation coefficients above 0.15 in AD brain, when the evidence for their association with AD was strong, suggesting the link between DNA methylation and AD. In conclusion DNA methylation status is intimately associated with correlated gene expression, particularly in AD brain.

View full abstract

De Novo ligand design is an automatic fragment-based design of molecules within a protein binding site of a known structure. A Bayesian Optimization Algorithm (BOA), a meta-heuristic algorithm, is introduced to join predocked fragments with a user-supplied list of fragments. A novel feature proposed is the simultaneous optimization of force field energy and a term enforcing 3D-overlap to known binding mode(s). The performance of the algorithm is tested on Liver X receptors (LXRs) using a library of about 14, 000 fragments and the binding mode of a known heterocyclic phenyl acetic acid to bias the design. We further introduce the use of GPU (Graphics Processing Unit) to overcome the excessive time required in evaluating each possible fragment combination. We show how the GPU utilization enables experimenting larger fragment sets and target receptors for more complex instances. The results show how the nVidia's Tesla C2050 GPU was utilized to enable the generation of complex agonists effectively. In fact, eight of the 1, 809 molecules designed for LXRs are found in the ZINC database of commercially available compounds.

View full abstract

The functional sites of multidomain proteins are often found at the interfaces of two or more domains. Therefore, the spatial arrangement of the domains is essential in understanding the functional mechanisms of multidomain proteins. However, an experimental determination of the whole structure of a multidomain protein is often difficult due to flexibility in inter-domain arrangement. We have developed a score function, named DINE, to detect probable docking poses generated in a rigid-body docking simulation. This score function takes into account the binding energy, information about the domain interfaces of homologous proteins, and the end-to-end distance spanned by the domain linker. We have examined the performance of DINE on 55 non-redundant known structures of two-domain proteins. In the results, the near-native docking poses were scored within the top 10 in 65.5% of the test cases. DINE scored the near-native poses higher in comparison with an existing domain assembly method, which also used binding energy and linker distance restraints. The results demonstrate that the domain-interface restraints of DINE are quite efficient in selecting near-native domain assemblies.

View full abstract

Alternative splicing plays an important role in eukaryotic gene expression by producing diverse proteins from a single gene. Predicting how genes are transcribed is of great biological interest. To this end, massively parallel whole transcriptome sequencing, often referred to as RNA-Seq, is becoming widely used and is revolutionizing the cataloging isoforms using a vast number of short mRNA fragments called reads. Conventional RNA-Seq analysis methods typically align reads onto a reference genome (mapping) in order to capture the form of isoforms that each gene yields and how much of every isoform is expressed from an RNA-Seq dataset. However, a considerable number of reads cannot be mapped uniquely. Those so-called multireads that are mapped onto multiple locations due to short read length and analogous sequences inflate the uncertainty as to how genes are transcribed. This causes inaccurate gene expression estimations and leads to incorrect isoform prediction. To cope with this problem, we propose a method for isoform prediction by iterative mapping. The positions from which multireads originate can be estimated based on the information of expression levels, whereas quantification of isoform-level expression requires accurate mapping. These procedures are mutually dependent, and therefore remapping reads is essential. By iterating this cycle, our method estimates gene expression levels more precisely and hence improves predictions of alternative splicing. Our method simultaneously estimates isoform-level expressions by computing how many reads originate from each candidate isoform using an EM algorithm within a gene. To validate the effectiveness of the proposed method, we compared its performance with conventional methods using an RNA-Seq dataset derived from a human brain. The proposed method had a precision of 66.7% and outperformed conventional methods in terms of the isoform detection rate.

View full abstract

In this study, we propose a new method to predict three-way interactions among proteins based on correlation coefficient of protein expression profiles. Although three-way interactions have not been studied well, this kind of interactions are important to understand the system of life. Previous studies reported the three-way interactions that based on switching mechanisms, in which a property or an expression level of a protein switches the mechanism of interactions between other two proteins. In this paper, we proposed a new method to predict three-way interactions based on the model in which A and B work together to effect on the expression level of C. We present the algorithm to predict the combinations of three proteins that have the three-way interaction, and evaluate it using our real proteome data.

View full abstract