Drug–target interaction prediction by integrating multiview network data

Drug–target interaction (DTI) prediction is a challenging step in further drug repositioning, drug discovery and drug design. The advent of high-throughput technologies brings convenience to the development of DTI prediction methods. With the generation of a high number of data sets, many mathematical models and computational algorithms have been developed to identify the potential drug–target pairs. However, most existing methods are proposed based on the single view data. By integrating the drug and target data from different views, we aim to get more stable and accurate prediction results.In this paper, a multiview DTI prediction method based on clustering is proposed. We first introduce a model for single view drug–target data. The model is formulated as an optimization problem, which aims to identify the clusters in both drug similarity network and target protein similarity network, and at the same time make the clusters with more known DTIs be connected together. Then the model is extended to multiview network data by maximizing the consistency of the clusters in each view. An approximation method is proposed to solve the optimization problem. We apply the proposed algorithms to two views of data. Comparisons with some existing algorithms show that the multiview DTI prediction algorithm can produce more accurate predictions. For the considered data set, we finally predict 54 possible DTIs. From the similarity analysis of the drugs/targets, enrichment analysis of DTIs and genes in each cluster, it is shown that the predicted DTIs have a high possibility to be true.     

Discovering DNA methylation patterns for long non-coding RNAs associated with cancer subtypes

Despite growing evidence demonstrates that the long non-coding ribonucleic acids (lncRNAs) are critical modulators for cancers, the knowledge about the DNA methylation patterns of lncRNAs is quite limited. We develop a systematic analysis pipeline to discover DNA methylation patterns for lncRNAs across multiple cancer subtypes from probe, gene and network levels. By using The Cancer Genome Atlas (TCGA) breast cancer methylation data, the pipeline discovers various DNA methylation patterns for lncRNAs across four major subtypes such as luminal A, luminal B, her2-enriched as well as basal-like. On the probe and gene level, we find that both differentially methylated probes and lncRNAs are subtype specific, while the lncRNAs are not as specific as probes. On the network level, the pipeline constructs differential co-methylation lncRNA network for each subtype. Then, it identifies both subtype specific and common lncRNA modules by simultaneously analyzing multiple networks. We show that the lncRNAs in subtype specific and common modules differ greatly in terms of topological structure, sequence conservation as well as expression. Furthermore, the subtype specific lncRNA modules serve as biomarkers to improve significantly the accuracy of breast cancer subtypes prediction. Finally, the common lncRNA modules associate with survival time of patients, which is critical for cancer therapy.     

Method for determining the absolute number concentration of nanoparticles from electrospray sources

We have developed a simple, fast, and accurate method to measure the absolute number concentration of nanoparticles in solution. The method combines electrospray differential mobility analysis (ES-DMA) with a statistical analysis of droplet-induced oligomer formation. A key feature of the method is that it allows determination of the absolute number concentration of particles by knowing only the droplet size generated from a particular ES source, thereby eliminating the need for sample-specific calibration standards or detailed analysis of transport losses. The approach was validated by comparing the total number concentration of monodispersed Au nanoparticles determined by ES-DMA with UV/vis measurements. We also show that this approach is valid for protein molecules by quantifying the absolute number concentration of Rituxan monoclonal antibody in solution. The methodology is applicable for quantification of any electrospray process coupled to an analytical tool that can distinguish monomers from higher order oligomers. The only requirement is that the droplet size distribution be evaluated. For users only interested in implementation of the theory, we provide a section that summarizes the relevant formulas. This method eliminates the need for sample-specific calibration standards or detailed analysis of transport losses.     

Identification of critical genes in microarray experiments by a Neuro-Fuzzy approach

Gene expression profiling by microarray technology is usually difficult to interpret into a simpler pattern. One approach to resolve the complexity of gene expression profiles is the application of artificial neural networks (ANNs). A potential difficulty in this strategy, however, is that the non-linear nature of ANN makes it essentially a 'black-box' computation process. Addition of a fuzzy logic approach is useful because it can complement ANN by explicitly specifying membership function during computation. We employed a hybrid approach of neural network and fuzzy logic to further analyze a published microarray study of gene responses to eight bacteria in human macrophages. The original analysis by hierarchical clustering found common gene responses to all bacteria but did not address individual responses. Our method allowed exploration of the gene response of the host to individual bacterium. We implemented a two-layer, feed-forward neural network containing the principle of 'competitive learning' (i.e. 'winner-take-all'). The weights of the trained neural network were fed into a fuzzy logic inference system. A new measurement, called the impact rating (IR) was also introduced to explore the degree of importance of each gene. To assess the reliability of the IR value, a bootstrap re-sampling method was applied to the dataset and a confidence level for each IR was obtained. Our approach has successfully uncovered the unique features of host response to individual bacterium. Further, application of gene ontology (GO) annotation to the genes of high IR values in each response has suggested new biological pathways for individual host-pathogen interactions.     

A single-cell drug efflux assay in bacteria by using a directly accessible femtoliter droplet array

Active efflux of drugs, such as antibiotics, from a cell is one of the major mechanisms that cause multi-drug resistance in bacteria. Here we report a method to assess drug efflux activity in individual Escherichia coli cells enclosed and isolated in a directly accessible femtoliter droplet array with a fluorogenic compound. The inhibitory effect of a chemical compound on an exogenously expressed efflux pump system from pathogenic bacteria was easily detected at the single-cell level. We also present a proof-of-principle experiment to screen for the gene encoding a drug efflux pump by collecting individual droplets containing single cells in which the drug efflux activity was restored after introduction of the exogenous gene from pathogenic bacteria. Our approach will be a useful tool to screen novel pump inhibitors and efflux pump genes, and to overcome infectious diseases caused by multi-drug-resistant bacteria.     

Unified QSAR and network-based computational chemistry approach to antimicrobials, part 1: multispecies activity models for antifungals

There are many pathogen microbial species with very different antimicrobial drugs susceptibility. In this work, we selected pairs of antifungal drugs with similar/dissimilar species predicted-activity profile and represented it as a large network, which may be used to identify drugs with similar mechanism of action. Computational chemistry prediction of the biological activity based on quantitative structure-activity relationships (QSAR) susbtantially increases the potentialities of this kind of networks, avoiding time and resource-consuming experiments. Unfortunately, most QSAR models are unspecific or predict activity against only one species. To solve this problem we developed a multispecies QSAR classification model, in which the outputs were the inputs of the aforementioned network. Overall model classification accuracy was 87.0% (161/185 compounds) in training, 83.4% (50/61) in validation, and 83.7% for 288 additional antifungal compounds used to extend model validation for network construction. The network predicted has 59 nodes (compounds), 648 edges (pairs of compounds with similar activity), low coverage density d = 37.8%, and distribution more close to normal than to exponential. These results are more characteristic of a not-overestimated random network, clustering different drug mechanisms of actions, than of a less useful power law network with few mechanisms (network hubs).     

A statistical method to detect chromosomal regions with DNA copy number alterations using SNP-array-based CGH data

Single nucleotide polymorphism (SNP) arrays were used to detect chromosomal regions with DNA copy number alterations. Current statistical methods for microarray-based comparative genomic hybridization (array-CGH) analysis generally assume certain relationships among adjacent markers on the same chromosome, and these assumptions may be questionable. For an SNP-array-based CGH study, multiple normal reference SNP arrays were collected. In order to utilize these normal reference SNP arrays, we derived an empirical distribution of signal ratios for each SNP marker. With an assumed threshold value for the overall error rate control and the defined signal ratio ranges for chromosomal amplification and deletion, we proposed a procedure to identify chromosomal alteration regions based on several bootstrapped one-sample t-tests and the false discovery rate control. When we have multiple arrays for different individuals with the same disease, our method can also be used to detect SNP markers for chromosomal alteration regions that are common among these individuals. We applied our method to a published SNP array data set for breast carcinoma cell lines. For an individual with breast cancer, numerous chromosomal alteration regions were identified. Compared to results of previous studies, our method identified more chromosomal alteration regions, with some being implicated in the literature to harbor genes associated with breast cancer. For multiple cancer arrays, our results suggested the existence of common chromosomal alteration regions. However, a high proportion of false positives also indicated that genetic variations among different individuals with breast cancer can be present.     

A Paleogenomic Algorithm for Reconstruction of Ancient Operons from Complete Microbial Genome Sequences

Operons, or co-transcribed and co-regulated contiguous sets of genes, in microbial genomes are poorly conserved across different genomes due to gene fusion, deletion, duplication and other genome shuffling processes. The currently available genomes are the results of numerous reshuffling and acceptance iterations. We hypothesized that in ancient times, when life was more primitive, functionally related genes existed in close proximity and operated together as an operon to simplify regulation. As more sophisticated regulation mechanisms became available during evolution the genes forming an operon could be separated by the above mentioned processes. If gene shuffling is a random event, neighbor gene pairs are more likely to be preserved than distant gene pairs. Thus, if enough gene pairs can be identified, the original operon could be reconstructed by assembling the pairs. Here we propose a novel paleogenomic method to reconstruct present neighbor gene pairs into "ancient" operons that possibly existed at some point during evolution.     

Three-Dimensional Aggregated Spheroid Model of Hepatocellular Carcinoma Using a 96-Pillar/Well Plate

A common method of three-dimensional (3D) cell cultures is embedding single cells in Matrigel. Separated cells in Matrigel migrate or grow to form spheroids but lack cell-to-cell interaction, which causes difficulty or delay in forming mature spheroids. To address this issue, we proposed a 3D aggregated spheroid model (ASM) to create large single spheroids by aggregating cells in Matrigel attached to the surface of 96-pillar plates. Before gelling the Matrigel, we placed the pillar inserts into blank wells where gravity allowed the cells to gather at the curved end. In a drug screening assay, the ASM with Hepatocellular carcinoma (HCC) cell lines showed higher drug resistance compared to both a conventional spheroid model (CSM) and a two-dimensional (2D) cell culture model. With protein expression, cytokine activation, and penetration analysis, the ASM showed higher expression of cancer markers associated with proliferation (p-AKT, p-Erk), tight junction formation (Fibronectin, ZO-1, Occludin), and epithelial cell identity (E-cadherin) in HCC cells. Furthermore, cytokine factors were increased, which were associated with immune cell recruitment/activation (MIF-3α), extracellular matrix regulation (TIMP-2), cancer interaction (IL-8, TGF-β2), and angiogenesis regulation (VEGF-A). Compared to CSM, the ASM also showed limited drug penetration in doxorubicin, which appears in tissues in vivo. Thus, the proposed ASM better recapitulated the tumor microenvironment and can provide for more instructive data during in vitro drug screening assays of tumor cells and improved prediction of efficacious drugs in HCC patients.     

Variable selection using probability density function similarity for support vector machine classification of high-dimensional microarray data

One problem with discriminant analysis of microarray data is representation of each sample by a large number of genes that are possibly irrelevant, insignificant or redundant. Methods of variable selection are, therefore, of great significance in microarray data analysis. To circumvent the problem, a new gene mining approach is proposed based on the similarity between probability density functions on each gene for the class of interest with respect to the others. This method allows the ascertainment of significant genes that are informative for discriminating each individual class rather than maximizing the separability of all classes. Then one can select genes containing important information about the particular subtypes of diseases. Based on the mined significant genes for individual classes, a support vector machine with local kernel transform is constructed for the classification of different diseases. The combination of the gene mining approach with support vector machine is demonstrated for cancer classification using two public data sets. The results reveal that significant genes are identified for each cancer, and the classification model shows satisfactory performance in training and prediction for both data sets.     

A closer look at "social" boundary genes reveals knowledge to gene expression profiles

As social network analysis is gaining popularity in modeling real world problems, the task of applying the social network model concepts and notions to biological data is still one of the most attractive research problems to be addressed. According, our work described in this paper focuses on a particular set of genes that reside on the community boundaries in gene co-expression networks. Stemmed from community mining problem in social networks, peripheries of communities (i.e., boundaries) can be used to aid certain biological analysis. The proposed method consists of three parts: 1) Finding communities of gene co-expression networks through clustering. 2) Analyzing stability of community structures by Monte Carlo method. 3) Designing of dynamic adoption of boundaries using geometric convexity. We validated our findings using breast cancer gene expression data from various studies. Our approach contributes to the new branch of applying social network mechanisms in biological data analysis, leading to new data mining strategies implied by witnessing social behaviors in gene expression analysis.     

Node-based differential network analysis in genomics

Gene dependency networks often undergo changes in response to different conditions. Understanding how these networks change across two conditions is an important task in genomics research. Most previous differential network analysis approaches assume that the difference between two condition-specific networks is driven by individual edges. Thus, they may fail in detecting key players which might represent important genes whose mutations drive the change of network. In this work, we develop a node-based differential network analysis (N-DNA) model to directly estimate the differential network that is driven by certain hub nodes. We model each condition-specific gene network as a precision matrix and the differential network as the difference between two precision matrices. Then we formulate a convex optimization problem to infer the differential network by combing a D-trace loss function and a row-column overlap norm penalty function. Simulation studies demonstrate that N-DNA provides more accurate estimate of the differential network than previous competing approaches. We apply N-DNA to ovarian cancer and breast cancer gene expression data. The model rediscovers known cancer-related genes and contains interesting predictions.     

Application of the network model for studying the delivery of colloidal drugs

From the viewpoint of colloidal transport phenomena, the individual drug particle’s movement through the tumor is studied by using the square network model. The effects of drug concentration, arterial pressure, and interstitial pressure of the tumor on the concentration breakthrough moment of drug particles are investigated by using the Brownian dynamics simulation method. From the simulation results, we find that the present network model and the Brownian dynamics simulation method can successfully analysis the therapeutic ranges achieved by the drug particles in the tumor at different magnitudes of the drug concentration, the inlet blood velocity, the arterial pressures, and the interstitial pressures of the tumor. Because of the high interstitial fluid pressure, the drug particles can only reach the peripheral region and cannot penetrate inside the target tumor efficiently, and the decreased interstitial fluid pressure will lead more drug particles to penetrate inside the tumor.     

Tumor stratification by a novel graph-regularized bi-clique finding algorithm

Due to involved disease mechanisms, many complex diseases such as cancer, demonstrate significant heterogeneity with varying behaviors, including different survival time, treatment responses, and recurrence rates. The aim of tumor stratification is to identify disease subtypes, which is an important first step towards precision medicine. Recent advances in profiling a large number of molecular variables such as in The Cancer Genome Atlas (TCGA), have enabled researchers to implement computational methods, including traditional clustering and bi-clustering algorithms, to systematically analyze high-throughput molecular measurements to identify tumor subtypes as well as their corresponding associated biomarkers.In this study we discuss critical issues and challenges in existing computational approaches for tumor stratification. We show that the problem can be formulated as finding densely connected sub-graphs (bi-cliques) in a bipartite graph representation of genomic data. We propose a novel algorithm that takes advantage of prior biology knowledge through a gene–gene interaction network to find such sub-graphs, which helps simultaneously identify both tumor subtypes and their corresponding genetic markers. Our experimental results show that our proposed method outperforms current state-of-the-art methods for tumor stratification.