International Journal of Computational Biology and Drug Design最新文献

Protein coding gene prediction by computational approaches is a fundamental step for genome annotation. However, it is a challenge to accurately predict eukaryotic genes in silico. By surveying the model genomes, we found that the Spearman's rank correlation coefficient between the number of experimental-verified genes and the size of genomes was 0.96 for all eukaryotes except plants, indicating the relationship between genome size and the number of coding genes can be expressed with a monotonic function. Regression analysis found that the relationship of total protein coding genes over genome size followed a logarithmic equation. We integrated the equation into ab initio gene prediction software to guide the gene prediction by constraining the total number of predicted genes. We evaluated the software in three eukaryotic genomes. Results showed that >90% of false positive predictions were removed while >80% of true positives were retained, resulting in much higher specificity.

Since co-expressed genes often are co-regulated by a group of transcription factors, different conditions (e.g. disease versus normal) may lead to different transcription factor activities and therefore different co-expression networks. We propose a method for identifying condition-specific co-expression networks by combining our recently developed network quasi-clique mining algorithm and the expected conditional F-statistic. We apply this method to compare the transcriptional programmes between the non-basal and basal types of breast cancers. The results provide a new perspective for studying gene interaction dynamics in cancers and assessing the effects of perturbation on key genes such as transcription factors. Our work is a way for dynamically characterising the gene interaction networks.

Protein sub-similarity matching remains largely unknown even though it is becoming one of the most important open problems in bioinformatics for drug and vaccine design. Variations in human immune responses to vaccines are, and thus responses, fail. We propose a new matching and protein alignment method based on clustering and Longest Common Subsequence (LCS) techniques. After clustering, we found LCS between a candidate protein and meningitis outer membrane antigen for each candidate. Each similarity was scored, and closest similarities were determined with statistical methods. We located three closely matching proteins among a total of 50 human immune system proteins. Moreover, we selected a HIV-1 related protein from one of scenarios, because it revealed a relationship between HIV and meningitis patients. We also found that Ω main chain torsion angle for atoms CA, C and N is the best angle for determining sub-similarities between meningitis antigen and immune proteins.

Software tools for the flexible examination of genomic sequence information derived from populations of organisms in a geospatial context are few in number, closely tied to Web-based resources, generally focused on one or a few loci or haplotypes, and typically produce a global phylogeny as a summary of relatedness. We sought instead to produce a portable, self-contained analysis tool that is efficiently focused on a geospatial display of specifically chosen polymorphism frequencies or combination frequencies from very large data sets of genome-scale sequence from multiple individuals. PolyLens is a Java-based, integral visual analytical toolkit which can systematically process population genomic data, visualise geographic distributions of genealogical lineages, and display allele distribution patterns. PolyLens is designed for users to visualise specific DNA sequences within each individual and its related location information in the existing data set.

In this paper, we present a framework for functional MR image restoration based on the Hidden Markov Tree (HMT) model. Under this scheme, the wavelet/contourlet coefficients of the distorted image are filtered using the HMT model of the baseline image to minimise the statistical divergence between two images. An iterative algorithm between image registration and HMT filtering is developed to achieve a trade-off between the least mean square error (in the spatial domain) and the minimum statistical divergence (in the spectral domain). We demonstrate that the proposed method can eliminate the motion artefacts (such as spikes and burring) in the Functional MR Imaging data more effectively, leading to reliable neural activity detection. This method can also be used for image restoration in other medical imaging applications.

The goal of this study is to develop a synthesisable computer-simulated model that mimics the function of a simplified renal system. Hardware description language has been used to simulate the model. In future phase of this study, the model will be realised on an electronic chip such as 'Field Programmable Gate Arrays'. The simulated model introduces a dynamic representation of the human body fluid balance under normal conditions and displays the change of urine flow with the amount of ingested water. The inputs of the model are average values of parameters extracted from the renal system. Some of these parameters and variables are: arterial pressure, daily ingested fluid volume, daily ingested sodium, daily ingested potassium, extracellular fluid volume, intracellular fluid volume, renin concentration, angiotensin II concentration, and aldosterone concentration. Our results show that the output of the model is in agreement with those of the literatures.

Next generation sequencing technology has matured, and with its current affordability, will replace the SNP chip as the genotyping tool of choice. Even with the current affordability of NGS, large scale studies will require careful study design to reduce cost. In this study, we designed an experiment to assess the accuracy of allele frequency estimated from pooled sequencing data. We compared the allele frequency estimated from sequencing data with the allele frequency estimated from individual SNP chip data and observed high correlations between them. However, by calculating error rate, we found that many SNPs had their allele frequency estimated from sequencing data significantly different from allele frequency estimated from SNP chip data. In conclusion, we found correlation is not an ideal measurement for comparing allele frequencies. And for the purpose of estimating allele frequency, we do not recommend using pooling with NGS as a cheaper alternative to genotype each sample individually.

In the present study both classification and correlation techniques have been successfully employed for the development of the models of diverse nature for the prediction of melanocortin 4-receptor (MC4 R) agonist activity using a dataset comprising of 56 analogues of 4-substituted piperidine-4-ol derivatives. Decision tree (DT), random forest (RF), moving average analysis (MAA) and multiple linear regression (MLR) were utilised for development of the said models. The statistical significance of models was assessed through specificity, sensitivity, overall accuracy, Mathew's correlation coefficient (MCC) and intercorrelation analysis. High accuracy of prediction up to 98% was observed using these models. Proposed models offer vast potential for providing lead structures for the development of potent therapeutic agents for the treatment of male sexual dysfunction.

Feature Selection (FS) is a process, which attempts to select features, which are more informative. In the supervised FS methods various feature subsets are evaluated using an evaluation function or metric to select only those features, which are related to the decision classes of the data under consideration. However, for many data mining applications, decision class labels are often unknown or incomplete, thus indicating the significance of unsupervised FS. However, in unsupervised learning, decision class labels are not provided. The problem is that not all features are important. Some of the features may be redundant, and others may be irrelevant and noisy. In this paper, a novel unsupervised FS in mammogram image, using rough set-based entropy measures, is proposed. A typical mammogram image processing system generally consists of mammogram image acquisition, pre-processing of image, segmentation, features extracted from the segmented mammogram image. The proposed method is used to select features from data set, the method is compared with the existing rough set-based supervised FS methods and classification performance of both methods are recorded and demonstrates the efficiency of the method.

Dysregulation of Pim-1 kinase has been implicated in several human cancers. Many potential inhibitors of PIM kinase have been reported, but potential bioactive compounds are still far from reach. Keeping this in mind, we have selected structurally known diverse Pim-1 kinase inhibitors to find novel small molecule drug-leads. A ligand-based pharmacophore model for Pim-1 kinase was developed using PHASE software. A four feature pharmacophoric hypothesis (AAHR) was used to develop atom-based 3D-QSAR model with the best regression coefficient of 0.9433 and Pearson-R of 0.9344. Compounds from Asinex platinum database were obtained whose pIC(50) values matched the 3D-QSAR model. Structural and molecular interaction studies on the training and test sets suggest that designing novel compounds hydrogen bond with Asp128 in the bioactive region of Pim-1 kinase would result in therapeutic success.