Yen-Jen Oyang

Relevant term suggestion in interactive web search based on contextual information in query session logs

This paper proposes an effective term suggestion approach to interactive Web search. Conventional approaches to making term suggestions involve extracting co-occurring keyterms from highly ranked retrieved documents. Such approaches must deal with term extraction difficulties and interference from irrelevant documents, and, more importantly, have difficulty extracting terms that are conceptually related but do not frequently co-occur in documents. In this paper, we present a new, effective log-based approach to relevant term extraction and term suggestion. Using this approach, the relevant terms suggested for a user query are those that co-occur in similar query sessions from search engine logs, rather than in the retrieved documents. In addition, the suggested terms in each interactive search step can be organized according to its relevance to the entire query session, rather than to the most recent single query as in conventional approaches. The proposed approach was tested using a proxy server log containing about two million query transactions submitted to search engines in Taiwan. The obtained experimental results show that the proposed approach can provide organized and highly relevant terms, and can exploit the contextual information in a users query session to make more effective suggestions.

Data classification with radial basis function networks based on a novel kernel density estimation algorithm

This work presents a novel learning algorithm for efficient construction of the radial basis function (RBF) networks that can deliver the same level of accuracy as the support vector machines (SVMs) in data classification applications. The proposed learning algorithm works by constructing one RBF subnetwork to approximate the probability density function of each class of objects in the training data set. With respect to algorithm design, the main distinction of the proposed learning algorithm is the novel kernel density estimation algorithm that features an average time complexity of O(nlogn), where n is the number of samples in the training data set. One important advantage of the proposed learning algorithm, in comparison with the SVM, is that the proposed learning algorithm generally takes far less time to construct a data classifier with an optimized parameter setting. This feature is of significance for many contemporary applications, in particular, for those applications in which new objects are continuously added into an already large database. Another desirable feature of the proposed learning algorithm is that the RBF networks constructed are capable of carrying out data classification with more than two classes of objects in one single run. In other words, unlike with the SVM, there is no need to resort to mechanisms such as one-against-one or one-against-all for handling datasets with more than two classes of objects. The comparison with SVM is of particular interest, because it has been shown in a number of recent studies that SVM generally are able to deliver higher classification accuracy than the other existing data classification algorithms. As the proposed learning algorithm is instance-based, the data reduction issue is also addressed in this paper. One interesting observation in this regard is that, for all three data sets used in data reduction experiments, the number of training samples remaining after a na/spl inodot//spl uml/ve data reduction mechanism is applied is quite close to the number of support vectors identified by the SVM software. This paper also compares the performance of the RBF networks constructed with the proposed learning algorithm and those constructed with a conventional cluster-based learning algorithm. The most interesting observation learned is that, with respect to data classification, the distributions of training samples near the boundaries between different classes of objects carry more crucial information than the distributions of samples in the inner parts of the clusters.

Risk of dementia after anaesthesia and surgery

BACKGROUND The potential relationship between anaesthesia, surgery and onset of dementia remains elusive. AIMS To determine whether the risk of dementia increases after surgery with anaesthesia, and to evaluate possible associations among age, mode of anaesthesia, type of surgery and risk of dementia. METHOD The study cohort comprised patients aged 50 years and older who were anaesthetised for the first time since 1995 between 1 January 2004 and 31 December 2007, and a control group of randomly selected patients matched for age and gender. Patients were followed until 31 December 2010 to identify the emergence of dementia. RESULTS Relative to the control group, patients who underwent anaesthesia and surgery exhibited an increased risk of dementia (hazard ratio = 1.99) and a reduced mean interval to dementia diagnosis. The risk of dementia increased in patients who received intravenous or intramuscular anaesthesia, regional anaesthesia and general anaesthesia. CONCLUSIONS The results of our nationwide, population-based study suggest that patients who undergo anaesthesia and surgery may be at increased risk of dementia.

Risk of Dementia in Patients with Insomnia and Long-term Use of Hypnotics: A Population-based Retrospective Cohort Study

Background Hypnotics have been reported to be associated with dementia. However, the relationship between insomnia, hypnotics and dementia is still controversial. We sought to examine the risk of dementia in patients with long-term insomnia and the contribution of hypnotics. Methods Data was collected from Taiwan’s Longitudinal Health Insurance Database. The study cohort comprised all patients aged 50 years or older with a first diagnosis of insomnia from 2002 to 2007. The comparison cohort consisted of randomly selected patients matched by age and gender. Each patient was individually tracked for 3 years from their insomnia index date to identify whether the patient had a first diagnosis of dementia. Cox regression was used to estimate hazard ratios (HRs) and 95% confidence intervals (CIs). Results We identified 5693 subjects with long-term insomnia and 28,465 individuals without. After adjusting for hypertension, diabetes mellitus, hyperlipidemia, and stroke, those with long-term insomnia had significantly higher risks of dementia (HR, 2.34; 95% CI, 1.92–2.85). Patients with long-term insomnia and aged 50 to 65 years had a higher increased risk of dementia (HR, 5.22; 95% CI, 2.62–10.41) than those older than 65 years (HR, 2.33; 95% CI, 1.90–2.88). The use of hypnotics with a longer half-life and at a higher prescribed dose predicted a greater increased risk of dementia. Conclusions Patients with long-term use of hypnotics have more than a 2-fold increased risk of dementia, especially those aged 50 to 65 years. In addition, the dosage and half-lives of the hypnotics used should be considered, because greater exposure to these medications leads to a higher risk of developing dementia.

Crosstalk between transcription factors and microRNAs in human protein interaction network

BackgroundGene regulatory networks control the global gene expression and the dynamics of protein output in living cells. In multicellular organisms, transcription factors and microRNAs are the major families of gene regulators. Recent studies have suggested that these two kinds of regulators share similar regulatory logics and participate in cooperative activities in the gene regulatory network; however, their combinational regulatory effects and preferences on the protein interaction network remain unclear.MethodsIn this study, we constructed a global human gene regulatory network comprising both transcriptional and post-transcriptional regulatory relationships, and integrated the protein interactome into this network. We then screened the integrated network for four types of regulatory motifs: single-regulation, co-regulation, crosstalk, and independent, and investigated their topological properties in the protein interaction network.ResultsAmong the four types of network motifs, the crosstalk was found to have the most enriched protein-protein interactions in their downstream regulatory targets. The topological properties of these motifs also revealed that they target crucial proteins in the protein interaction network and may serve important roles of biological functions.ConclusionsAltogether, these results reveal the combinatorial regulatory patterns of transcription factors and microRNAs on the protein interactome, and provide further evidence to suggest the connection between gene regulatory network and protein interaction network.

Dynamic functional modules in co-expressed protein interaction networks of dilated cardiomyopathy

BackgroundMolecular networks represent the backbone of molecular activity within cells and provide opportunities for understanding the mechanism of diseases. While protein-protein interaction data constitute static network maps, integration of condition-specific co-expression information provides clues to the dynamic features of these networks. Dilated cardiomyopathy is a leading cause of heart failure. Although previous studies have identified putative biomarkers or therapeutic targets for heart failure, the underlying molecular mechanism of dilated cardiomyopathy remains unclear.ResultsWe developed a network-based comparative analysis approach that integrates protein-protein interactions with gene expression profiles and biological function annotations to reveal dynamic functional modules under different biological states. We found that hub proteins in condition-specific co-expressed protein interaction networks tended to be differentially expressed between biological states. Applying this method to a cohort of heart failure patients, we identified two functional modules that significantly emerged from the interaction networks. The dynamics of these modules between normal and disease states further suggest a potential molecular model of dilated cardiomyopathy.ConclusionsWe propose a novel framework to analyze the interaction networks in different biological states. It successfully reveals network modules closely related to heart failure; more importantly, these network dynamics provide new insights into the cause of dilated cardiomyopathy. The revealed molecular modules might be used as potential drug targets and provide new directions for heart failure therapy.

A multimedia storage system for on-demand playback

This paper presents the design of a multimedia storage system for on-demand playback. The design stresses effective utilization of disk bandwidth with minimal data buffer to minimize overall system costs. The design procedure is most distinctive in the following two aspects: it bases on a tight upper bound of the lumped disk seek time for the Scan disk scheduling algorithm to achieve effective utilization of the disk bandwidth; and it starts with a general two-level hierarchical disk array structure to derive the optimal configuration for specific requirements. >

ProteDNA: a sequence-based predictor of sequence-specific DNA-binding residues in transcription factors

This article presents the design of a sequence-based predictor named ProteDNA for identifying the sequence-specific binding residues in a transcription factor (TF). Concerning protein–DNA interactions, there are two types of binding mechanisms involved, namely sequence-specific binding and nonspecific binding. Sequence-specific bindings occur between protein sidechains and nucleotide bases and correspond to sequence-specific recognition of genes. Therefore, sequence-specific bindings are essential for correct gene regulation. In this respect, ProteDNA is distinctive since it has been designed to identify sequence-specific binding residues. In order to accommodate users with different application needs, ProteDNA has been designed to operate under two modes, namely, the high-precision mode and the balanced mode. According to the experiments reported in this article, under the high-precision mode, ProteDNA has been able to deliver precision of 82.3%, specificity of 99.3%, sensitivity of 49.8% and accuracy of 96.5%. Meanwhile, under the balanced mode, ProteDNA has been able to deliver precision of 60.8%, specificity of 97.6%, sensitivity of 60.7% and accuracy of 95.4%. ProteDNA is available at the following websites: http://protedna.csbb.ntu.edu.tw/ http://protedna.csie.ntu.edu.tw/ http://bio222.esoe.ntu.edu.tw/ProteDNA/.

An Incremental Hierarchical Data Clustering Algorithm Based on Gravity Theory

One of the main challenges in the design of modern clustering algorithms is that, in many applications, new data sets are continuously added into an already huge database. As a result, it is impractical to carry out data clustering from scratch whenever there are new data instances added into the database. One way to tackle this challenge is to incorporate a clustering algorithm that operates incrementally. Another desirable feature of clustering algorithms is that a clustering dendrogram is generated. This feature is crucial for many applications in biological, social, and behavior studies, due to the need to construct taxonomies. This paper presents the GRIN algorithm, an incremental hierarchical clustering algorithm for numerical data sets based on gravity theory in physics. The GRIN algorithm delivers favorite clustering quality and generally features O(n) time complexity. One main factor that makes the GRIN algorithm be able to deliver favorite clustering quality is that the optimal parameters settings in the GRIN algorithm are not sensitive to the distribution of the data set. On the other hand, many modern clustering algorithms suffer unreliable or poor clustering quality when the data set contains highly skewed local distributions so that no optimal values can be found for some global parameters. This paper also reports the experiments conducted to study the characteristics of the GRIN algorithm.

MicroRNA-Regulated Protein-Protein Interaction Networks and Their Functions in Breast Cancer

MicroRNAs, which are small endogenous RNA regulators, have been associated with various types of cancer. Breast cancer is a major health threat for women worldwide. Many miRNAs were reported to be associated with the progression and carcinogenesis of breast cancer. In this study, we aimed to discover novel breast cancer-related miRNAs and to elucidate their functions. First, we identified confident miRNA-target pairs by combining data from miRNA target prediction databases and expression profiles of miRNA and mRNA. Then, miRNA-regulated protein interaction networks (PINs) were constructed with confident pairs and known interaction data in the human protein reference database (HPRD). Finally, the functions of miRNA-regulated PINs were elucidated by functional enrichment analysis. From the results, we identified some previously reported breast cancer-related miRNAs and functions of the PINs, e.g., miR-125b, miR-125a, miR-21, and miR-497. Some novel miRNAs without known association to breast cancer were also found, and the putative functions of their PINs were also elucidated. These include miR-139 and miR-383. Furthermore, we validated our results by receiver operating characteristic (ROC) curve analysis using our miRNA expression profile data, gene expression-based outcome for breast cancer online (GOBO) survival analysis, and a literature search. Our results may provide new insights for research in breast cancer-associated miRNAs.