• Journal of Information Processing Systems
• /
• 제8권4호
• /
• pp.575-588
• /
• 2012

Exploiting unlabeled text data with a relatively small labeled corpus has been an active and challenging research topic in text mining, due to the recent growth of the amount of biomedical literature. Biomedical named-entity recognition is an essential prerequisite task before effective text mining of biomedical literature can begin. This paper proposes an Active Co-Training (ACT) algorithm for biomedical named-entity recognition. ACT is a semi-supervised learning method in which two classifiers based on two different feature sets iteratively learn from informative examples that have been queried from the unlabeled data. We design a new classification problem to measure the informativeness of an example in unlabeled data. In this classification problem, the examples are classified based on a joint view of a feature set to be informative/non-informative to both classifiers. To form the training data for the classification problem, we adopt a query-by-committee method. Therefore, in the ACT, both classifiers are considered to be one committee, which is used on the labeled data to give the informativeness label to each example. The ACT method outperforms the traditional co-training algorithm in terms of f-measure as well as the number of training iterations performed to build a good classification model. The proposed method tends to efficiently exploit a large amount of unlabeled data by selecting a small number of examples having not only useful information but also a comprehensive pattern.

• 대한원격탐사학회지
• /
• 제23권5호
• /
• pp.393-400
• /
• 2007

We present in this paper a novel power load prediction method using temporal pattern mining from AMR(Automatic Meter Reading) data. Since the power load patterns have time-varying characteristic and very different patterns according to the hour, time, day and week and so on, it gives rise to the uninformative results if only traditional data mining is used. Also, research on data mining for analyzing electric load patterns focused on cluster analysis and classification methods. However despite the usefulness of rules that include temporal dimension and the fact that the AMR data has temporal attribute, the above methods were limited in static pattern extraction and did not consider temporal attributes. Therefore, we propose a new classification method for predicting power load patterns. The main tasks include clustering method and temporal classification method. Cluster analysis is used to create load pattern classes and the representative load profiles for each class. Next, the classification method uses representative load profiles to build a classifier able to assign different load patterns to the existing classes. The proposed classification method is the Calendar-based temporal mining and it discovers electric load patterns in multiple time granularities. Lastly, we show that the proposed method used AMR data and discovered more interest patterns.

• 한국생물정보학회:학술대회논문집
• /
• 한국생물정보시스템생물학회 2006년도 Principles and Practice of Microarray for Biomedical Researchers
• /
• pp.30-36
• /
• 2006

Tissue microarray (TMA) is an array-based technology allowing the examination of hundreds of tissue samples on a single slide. To handle, exchange, and disseminate TMA data, we need standard representations of the methods used, of the data generated, and of the clinical and histopathological information related to TMA data analysis. This study aims to create a comprehensive data model with flexibility that supports diverse experimental designs and with expressivity and extensibility that enables an adequate and comprehensive description of new clinical and histopathological data elements. We designed a Tissue Microarray Object Model (TMA-OM). Both the Array Information and the Experimental Procedure models are created by referring to Microarray Gene Expression Object Model, Minimum Information Specification For In Situ Hybridization and Immunohistochemistry Experiments (MISFISHIE), and the TMA Data Exchange Specifications (TMA DES). The Clinical and Histopathological Information model is created by using CAP Cancer Protocols and National Cancer Institute Common Data Elements (NCI CDEs). MGED Ontology, UMLS and the terms extracted from CAP Cancer Protocols and NCI CDEs are used to create a controlled vocabulary for unambiguous annotation. We implemented a web-based application for TMA-OM, supporting data export in XML format conforming to the TMA DES or the DTD derived from TMA-OM. TMA-OM provides a comprehensive data model for storage, analysis and exchange of TMA data and facilitates model-level integration of other biological models.

• 대한예방의학회:학술대회논문집
• /
• 대한예방의학회 2002년도 하계워크샵 연제집
• /
• pp.76-88
• /
• 2002

• 한국생물정보학회:학술대회논문집
• /
• 한국생물정보시스템생물학회 2005년도 BIOINFO 2005
• /
• pp.313-318
• /
• 2005

Estimating the reliability of protein-protein interaction data sets obtained by high-throughput technologies such as yeast two-hybrid assays and mass spectrometry is of great importance. We develop a maximum likelihood estimation method that uses both protein localization and gene expression data to estimate the reliability of protein interaction data sets. By integrating protein localization data and gene expression data, we can obtain more accurate estimates of the reliability of various interaction data sets. We apply the method to protein physical interaction data sets and protein complex data sets. The reliability of the yeast two-hybrid interactions by Ito et al. (2001) is 27%, and that by Uetz et at.(2000) is 68%. The reliability of the protein complex data sets using tandem affinity purification-mass spec-trometry (TAP) by Gavin et at. (2002) is 45%, and that using high-throughput mass spectrometric protein complex identification (HMS-PCI) by Ho et al. (2002) is 20%. The method is general and can be applied to analyze any protein interaction data sets.

• 한국생물정보학회:학술대회논문집
• /
• 한국생물정보시스템생물학회 2005년도 BIOINFO 2005
• /
• pp.357-360
• /
• 2005

In this paper we consider the well-known semiparametric proportional hazards (PH) models for survival analysis. These models are usually used with few covariates and many observations (subjects). But, for a typical setting of gene expression data from DNA microarray, we need to consider the case where the number of covariates p exceeds the number of samples n. For a given vector of response values which are times to event (death or censored times) and p gene expressions (covariates), we address the issue of how to reduce the dimension by selecting the significant genes. This approach enable us to estimate the survival curve when n < < p. In our approach, rather than fixing the number of selected genes, we will assign a prior distribution to this number. The approach creates additional flexibility by allowing the imposition of constraints, such as bounding the dimension via a prior, which in effect works as a penalty. To implement our methodology, we use a Markov Chain Monte Carlo (MCMC) method. We demonstrate the use of the methodology to diffuse large B-cell lymphoma (DLBCL) complementary DNA(cDNA) data.

• 한국생물정보학회:학술대회논문집
• /
• 한국생물정보시스템생물학회 2005년도 BIOINFO 2005
• /
• pp.237-242
• /
• 2005

Predicting the cellular location of an unknown protein gives a valuable information for inferring the possible function of the protein. For more accurate prediction system, we need a good feature extraction method that transforms the raw sequence data into the numerical feature vector, minimizing information loss. In this paper, we propose new methods of extracting underlying features only from the sequence data by computing pairwise sequence alignment scores. In addition, we use composition based features to improve prediction accuracy. To construct an SVM ensemble from separately trained SVM classifiers, we propose specificity based weighted majority voting. The overall prediction accuracy evaluated by the 5-fold cross-validation reached 88.53% for the eukaryotic animal data set. By comparing the prediction accuracy of various feature extraction methods, we could get the biological insight on the location of targeting information. Our numerical experiments confirm that our new feature extraction methods are very useful for predicting subcellular localization of proteins.

• 한국생물정보학회:학술대회논문집
• /
• 한국생물정보시스템생물학회 2005년도 BIOINFO 2005
• /
• pp.15-20
• /
• 2005

We developed a database system to enable efficient and high-throughput transposon analyses in rice. We grow large-scale mutant series of rice by taking advantage of an active MITE transposon mPing, and apply the transposon display method to them to study correlation between genotypes and phenotypes. But the analytical phase, in which we find mutation spots from waveform data called fragment profiles, involves several problems from a viewpoint of labor amount, data management, and reliability of the result. As a solution, our database system manages all the analytical data throughout the experiments, and provides several functions and well designed web interfaces to perform overall analyses reliably and efficiently.

• 한국정보과학회:학술대회논문집
• /
• 한국정보과학회 2008년도 한국컴퓨터종합학술대회논문집 Vol.35 No.1 (C)
• /
• pp.212-216
• /
• 2008

기존의 알려진 데이터에 기반하여 분류 알고리즘을 통해 새로운 생물학적인 사실을 예측하는 것은 생물학 연구에 매우 유용하다. 하지만 생물학 데이터 분류 문제에서 positive 데이터만 존재할 뿐, negative 데이터는 존재하지 않는 경우가 많다. 이와 같은 상황에서는 많은 경우에 임의로 negative data를 구성하여 사용하게 된다. 하지만, negative 데이터는 실제로 negative임이 보장된 것이 아니고, 임의로 생성된 데이터의 특성에 따라 분류 성능 및 모델의 특성에 많은 차이를 보일 수 있다. 따라서 본 논문에서는 단일 클래스 분류 알고리즘 중 하나인 support vector data description(SVDD) 방법을 이용하여 실제 microRNA target 예측 문제에서 positive 데이터만을 이용하여 학습하고 분류를 수행하였다. 이를 통해 일반적인 이진 분류 방법에 비해 이와 같은 방법이 실제 생물학 문제에 보다 적합하게 적용될 수 있음을 확인한다.

• 한국생물정보학회:학술대회논문집
• /
• 한국생물정보시스템생물학회 2003년도 제2차 연례학술대회 발표논문집
• /
• pp.1-1
• /
• 2003

It is clear that computers will play a key role in the biology of the future. Even now, it is virtually impossible to keep track of the key proteins, their names and associated gene names, physical constants(e.g. binding constants, reaction constants, etc.), and hewn physical and genetic interactions without computational assistance. In this sense, computers act as an auxiliary brain, allowing one to keep track of thousands of complex molecules and their interactions. With the advent of gene expression array technology, many experiments are simply impossible without this computer assistance. In the future, as we seek to integrate the reductionist description of life provided by genomic sequencing into complex and sophisticated models of living systems, computers will play an increasingly important role in both analyzing data and generating experimentally testable hypotheses. The future of bioinformatics is thus being driven by potent technological and scientific forces. On the technological side, new experimental technologies such as microarrays, protein arrays, high-throughput expression and three-dimensional structure determination prove rapidly increasing amounts of detailed experimental information on a genomic scale. On the computational side, faster computers, ubiquitous computing systems, high-speed networks provide a powerful but rapidly changing environment of potentially immense power. The challenges we face are enormous: How do we create stable data resources when both the science and computational technology change rapidly? How do integrate and synthesize information from many disparate subdisciplines, each with their own vocabulary and viewpoint? How do we 'liberate' the scientific literature so that it can be incorporated into electronic resources? How do we take advantage of advances in computing and networking to build the international infrastructure needed to support a complete understanding of biological systems. The seeds to the solutions of these problems exist, at least partially, today. These solutions emphasize ubiquitous high-speed computation, database interoperation, federation, and integration, and the development of research networks that capture scientific knowledge rather than just the ABCs of genomic sequence. 1 will discuss a number of these solutions, with examples from existing resources, as well as area where solutions do not currently exist with a view to defining what bioinformatics and biology will look like in the future.