Search PubMed⌕ Search

SEARCH · Search PubMed

Results for “unsupervised clustering”

Search indexed PubMed citations on genomics, clinical trials, systematic reviews and public health. Explore titles, authors and supplied subject terms, then open the PubMed record.

Quote a phrase for an exact phrase match. Source license links do not imply unrestricted reuse.

At least 577 records · Page 32Linked to original sources
PubMed2003

Correlations between pulp properties of eucalyptus clones and leaf volatiles using automated solid-phase microextraction.

Analysis of biogenic volatile organic compounds (BVOC) of 14 Eucalyptus clones has been performed using an automated headspace solid-phase microextraction (HS-SPME) coupled to gas chromatography (GC)/ion trap mass spectrometry (ITMS) method. Correlations between pulp properties of Eucalyptus clones and the BVOC of their leaf headspaces were studied. The compounds alpha-terpineol and the sesquiterpene beta-eudesmol were positively correlated with S5, a property related to the hemicelluose content in the pulp. Qualitative results obtained with automated HS-SPME were sufficient to group together the same species and related hybrids through cluster analysis and were confirmed through principal component analysis. A preliminary separation of the essential oils of Eucalyptus dunnii through comprehensive two-dimensional gas chromatography (GC x GC) showed approximately 580 peaks compared to approximately 60 in a typical GC/ITMS first-dimension chromatogram. The potential of HS-SPME coupled to GC x GC to improve the separation of Eucalyptus volatiles and other plant essential oils looks extremely promising for new applications of unsupervised learning methods.

Autoanalysis↗
PubMed2007

High-resolution (13)C nuclear magnetic resonance spectroscopy pattern recognition of fish oil capsules.

13C NMR (nuclear magnetic resonance) spectroscopy, in conjunction with multivariate analysis of commercial fish oil-related health food products, have been used to provide discrimination concerning the nature, composition, refinement, and/or adulteration or authentication of the products. Supervised (probabilistic neural networks, PNN) and unsupervised (principal component analysis, PCA; Kohonen neural networks; generative topographic mapping, GTM) pattern recognition techniques were used to visualize and classify samples. Simple PCA score plots demonstrated excellent, but not totally unambiguous, class distinctions, whereas Kohonen and GTM visualization provided better results. Quantitative class predictions with accuracies >95% were achieved with PNN analysis. Trout, salmon, and cod oils were completely and correctly classified. Samples reported to be salmon oils and cod liver oils did not cluster with true salmon and cod liver oil samples, indicating mislabeling or adulteration.

Animals↗
PubMed2001

Characterization of enzootic foci of Venezuelan equine encephalitis virus in western Venezuela.

The distribution of the sylvatic subtype ID Venezuelan equine encephalitis (VEE) viruses in the lowland tropical forests of western Venezuela was investigated using remote sensing and geographic information system technologies. Landsat 5 Thematic Mapper satellite imagery was used to study the reflectance patterns of VEE endemic foci and to identify other locations with similar reflectance patterns. Enzootic VEE virus variants isolated during this study are the closest genetic relatives of the epizootic viruses that emerged in western Venezuela during 1992-1993. VEE virus surveillance was conducted by exposing sentinel hamsters to mosquito bites and trapping wild vertebrates in seven forests identified and located by means of the satellite image. We isolated VEE viruses from 48 of a total of 1,363 sentinel hamsters in two of the forests on six occasions, in both dry and wet seasons. None of the 12 small vertebrates captured in 8,190 trap-nights showed signs of previous VEE virus infection. The satellite image was classified into 13 validated classes of land use/vegetation using unsupervised and supervised techniques. Data derived from the image consisted of the raw digital values of near- and mid-infrared bands 4, 5, and 7, derived Tasseled Cap indices of wetness, greenness, and brightness, and the Normalized Difference Vegetation Index. Digitized maps provided ancillary data of elevation and soil geomorphology. Image enhancement was applied using Principal Component Analysis. A digital layer of roads together with georeferenced images was used to locate the study sites. A cluster analysis using the above data revealed two main groups of dense forests separated by spectral properties, altitude, and soil geomorphology. Virus was isolated more frequently from the forest type identified on flat flood plains of main rivers rather than the forest type found on the rolling hills of the study area. The spatial analysis suggests that mosquitoes carrying the enzootic viruses would reach 82-97% of the total land area by flying only 1-3 km from forests. We hypothesize that humans within that area are at risk of severe disease caused by enzootic ID VEE viruses. By contrast, equines could actually become naturally vaccinated, thus preventing the local emergence of epizootic IC VEE virus strains and protecting humans indirectly.

Animals↗
PubMedDate not supplied

Analysis and classification of proton NMR spectra of lipoprotein fractions from healthy volunteers and patients with cancer or CHD.

Human blood plasma samples from 52 subjects were collected and the very low density lipoprotein (VLDL), intermediate density lipoprotein (IDL), low density lipoprotein (LDL) and high density lipoprotein were isolated by serial ultra centrifugation. 600 MHz 1H NMR spectra of the lipoprotein fractions were acquired. The methyl and methylene regions in the spectra of VLDL, LDL and HDL were utilised in further analyses via Kohonen neural networks (KNN) and generative topographic mapping (GTM), two related examples of (unsupervised learning) self-organising feature mapping techniques. Systematic variations in lipoprotein profiles can be substantially visualised through the use of KNN and GTM. The relationship between the sample positions in the Kohonen plot was visualised by surface plots of the corresponding VLDL and HDL cholesterol and VLDL triglyceride contents. The GTM maps of the VLDL and HDL fractions were used to investigate the individual properties of selected samples. A large number of the cancer patients were found clustered in the VLDL GTM map, and GTM map positions of samples in relation to CHD, diabetes and renal failure could be found. Although the study group here considered is heterogeneous in respect to age, sex, type of disease and medications within each defined class, classification of VLDL and HDL data with probabilistic neural network (PNN) was quite successful with respect to the groupings: cancer, CHD, volunteers and other (comprising patients with other diseases). Statistics based on 15 independent sets of PNN calculations gave true positive fractions usually higher than 0.83 and false positive fractions lower than 0.088. Attempts to use the corresponding LDL data and four classes were uniformly poor although some classifications (e.g., volunteer versus CHD) were easily performed.

Adult↗
PubMed2004

Estimating genomic coexpression networks using first-order conditional independence.

We describe a computationally efficient statistical framework for estimating networks of coexpressed genes. This framework exploits first-order conditional independence relationships among gene-expression measurements to estimate patterns of association. We use this approach to estimate a coexpression network from microarray gene-expression measurements from Saccharomyces cerevisiae. We demonstrate the biological utility of this approach by showing that a large number of metabolic pathways are coherently represented in the estimated network. We describe a complementary unsupervised graph search algorithm for discovering locally distinct subgraphs of a large weighted graph. We apply this algorithm to our coexpression network model and show that subgraphs found using this approach correspond to particular biological processes or contain representatives of distinct gene families.

Algorithms↗
PubMed2006

An efficient semi-unsupervised gene selection method via spectral biclustering.

Gene selection is an important issue in microarray data processing. In this paper, we propose an efficient method for selecting relevant genes. First, we use spectral biclustering to obtain the best two eigenvectors for class partition. Then gene combinations are selected based on the similarity between the genes and the best eigenvectors. We demonstrate our semi-unsupervised gene selection method using two microarray cancer data sets, i.e., the lymphoma and the liver cancer data sets, where our method is able to identify a single gene or a two-gene combinations which can lead to predictions with very high accuracy.

Artificial Intelligence↗
PubMed2004

Compliance with 14-day primaquine therapy for radical cure of vivax malaria--a randomized placebo-controlled trial comparing unsupervised with supervised treatment.

The only available treatment that can eliminate the latent hypnozoite reservoir of vivax malaria is a 14 d course of primaquine (PQ). A potential problem with long-course chemotherapy is the issue of compliance after clinical symptoms have subsided. The present study, carried out at an Afghan refugee camp in Pakistan, between June 2000 and August 2001, compared 14 d treatment in supervised and unsupervised groups in which compliance was monitored by comparison of relapse rates. Clinical cases recruited by passive case detection were randomised by family to placebo, supervised, or unsupervised groups, and treated with chloroquine (25 mg/kg) over 3 days to eliminate erythrocytic stages. Individuals with glucose-6-phosphate dehydrogenase (G6PD) deficiency were excluded from the trial. Cases allocated to supervision were given directly observed treatment (0.25 mg PQ/kg body weight) once per day for 14 days. Cases allocated to the unsupervised group were provided with 14 PQ doses upon enrollment and strongly advised to complete the course. A total of 595 cases were enrolled. After 9 months of follow up PQ proved equally protective against further episodes of P. vivax in supervised (odds ratio 0.35, 95% CI 0.21-0.57) and unsupervised (odds ratio 0.37, 95% CI 0.23-0.59) groups as compared to placebo. All age groups on supervised or unsupervised treatment showed a similar degree of protection even though the risk of relapse decreased with age. The study showed that a presumed problem of poor compliance may be overcome with simple health messages even when the majority of individuals are illiterate and without formal education. Unsupervised treatment with 14-day PQ when combined with simple instruction can avert a significant amount of the morbidity associated with relapse in populations where G6PD deficiency is either absent or readily diagnosable.

Adolescent↗
PubMed2007

Discovering biomarkers from gene expression data for predicting cancer subgroups using neural networks and relational fuzzy clustering.

BACKGROUND: The four heterogeneous childhood cancers, neuroblastoma, non-Hodgkin lymphoma, rhabdomyosarcoma, and Ewing sarcoma present a similar histology of small round blue cell tumor (SRBCT) and thus often leads to misdiagnosis. Identification of biomarkers for distinguishing these cancers is a well studied problem. Existing methods typically evaluate each gene separately and do not take into account the nonlinear interaction between genes and the tools that are used to design the diagnostic prediction system. Consequently, more genes are usually identified as necessary for prediction. We propose a general scheme for finding a small set of biomarkers to design a diagnostic system for accurate classification of the cancer subgroups. We use multilayer networks with online gene selection ability and relational fuzzy clustering to identify a small set of biomarkers for accurate classification of the training and blind test cases of a well studied data set. RESULTS: Our method discerned just seven biomarkers that precisely categorized the four subgroups of cancer both in training and blind samples. For the same problem, others suggested 19-94 genes. These seven biomarkers include three novel genes (NAB2, LSP1 and EHD1 - not identified by others) with distinct class-specific signatures and important role in cancer biology, including cellular proliferation, transendothelial migration and trafficking of MHC class antigens. Interestingly, NAB2 is downregulated in other tumors including Non-Hodgkin lymphoma and Neuroblastoma but we observed moderate to high upregulation in a few cases of Ewing sarcoma and Rabhdomyosarcoma, suggesting that NAB2 might be mutated in these tumors. These genes can discover the subgroups correctly with unsupervised learning, can differentiate non-SRBCT samples and they perform equally well with other machine learning tools including support vector machines. These biomarkers lead to four simple human interpretable rules for the diagnostic task. CONCLUSION: Although the proposed method is tested on a SRBCT data set, it is quite general and can be applied to other cancer data sets. Our scheme takes into account the interaction between genes as well as that between genes and the tool and thus is able find a very small set and can discover novel genes. Our findings suggest the possibility of developing specialized microarray chips or use of real-time qPCR assays or antibody based methods such as ELISA and western blot analysis for an easy and low cost diagnosis of the subgroups.

Biomarkers, Tumor↗
PubMed2006

Spike sorting: Bayesian clustering of non-stationary data.

Spike sorting involves clustering spikes recorded by a micro-electrode according to the source neurons. It is a complicated task, which requires much human labor, in part due to the non-stationary nature of the data. We propose to automate the clustering process in a Bayesian framework, with the source neurons modeled as a non-stationary mixture-of-Gaussians. At a first search stage, the data are divided into short time frames, and candidate descriptions of the data as mixtures-of-Gaussians are computed for each frame separately. At a second stage, transition probabilities between candidate mixtures are computed, and a globally optimal clustering solution is found as the maximum-a-posteriori solution of the resulting probabilistic model. The transition probabilities are computed using local stationarity assumptions, and are based on a Gaussian version of the Jensen-Shannon divergence. We employ synthetically generated spike data to illustrate the method and show that it outperforms other spike sorting methods in a non-stationary scenario. We then use real spike data and find high agreement of the method with expert human sorters in two modes of operation: a fully unsupervised and a semi-supervised mode. Thus, this method differs from other methods in two aspects: its ability to account for non-stationary data, and its close to human performance.

Algorithms↗
PubMedDate not supplied

Using self-organizing maps to identify potential halo white dwarfs.

We present the results of an unsupervised classification of the disk and halo white dwarf populations in the solar neighborhood. The classification is done by merging the results of detailed Monte Carlo (MC) simulations, which reproduce very well the characteristics of the white dwarf populations in the solar neighborhood, with a catalogue of real stars. The resulting composite catalogue is analyzed using a competitive learning algorithm. In particular we have used the so-called self-organized map. The MC simulated stars are used as tracers and help in identifying the resulting clusters. The results of such an strategy turn out to be quite satisfactory, suggesting that this approach can provide an useful framework for analyzing large databases of white dwarfs with well determined kinematical, spatial and photometric properties once they become available in the next decade. Moreover, the results are of astrophysical interest as well, since a straightforward interpretation of several recent astronomical observations, like the detected microlensing events in the direction of the Magellanic Clouds, the possible detection of high proper motion white dwarfs in the Hubble Deep Field and the discovery of high velocity white dwarfs in the solar neighborhood, suggests that a fraction of the baryonic dark matter component of our galaxy could be in the form of old and dim halo white dwarfs.

Artificial Intelligence↗
PubMed2006

Growing hierarchical tree SOM: an unsupervised neural network with dynamic topology.

In this paper we introduce a tree structured self-organizing network, called the Growing Hierarchical Tree SOM (GHTSOM), that combines unsupervised learning with a dynamic topology for hierarchical classification of unlabelled data sets. The main feature of the proposed model is a SOM-like self-organizing process that allows the network to adapt the topology of each layer of the hierarchy to the characteristics of the training set. In particular the self-organization is obtained in two steps: the first one concerns the learning phase and is finalized with the creation of a tree of SOMs, while the second one is in regard to the clustering phase and provides the formation of classes for each level of the tree (hence self-organization not only for training but also for the creation of topological connections). As a result the network works without the need for user-defined parameters. Experimental results are proposed on both synthetic and real data sets.

Algorithms↗
PubMed2006

A self-organizing CMAC network with gray credit assignment.

This paper attempts to incorporate the structure of the cerebellar-model-articulation-controller (CMAC) network into the Kohonen layer of the self-organizing map (SOM) to construct a self-organizing CMAC (SOCMAC) network. The proposed SOCMAC network can perform the function of an SOM and can distribute the learning error into the memory contents of all addressed hypercubes as a CMAC. The learning of the SOCMAC is in an unsupervised manner. The neighborhood region of the SOCMAC is implicit in the structure of a two-dimensional CMAC network and needs not be defined in advance. Based on gray relational analysis, a credit-assignment technique for SOCMAC learning is introduced to hasten the overall learning process. This paper also analyzes the convergence properties of the SOCMAC. It is shown that under the proposed updating rule, both the memory contents and the state outputs of the SOCMAC converge almost surely. The SOCMAC is applied to solve both data-clustering and data-classification problems, and simulation results show that the proposed network achieves better performance than other known SOMs.

Algorithms↗
PubMedDate not supplied

Artificial neural networks and three-dimensional digital morphology: a pilot study.

In this pilot study, we used an unsupervised learning algorithm for self-organization and pattern matching to create feature maps that can be applied to morphological problems. We designed a network to analyze 83 first and/or second upper and lower molar sets representing 13 anthropoid primate species, based on three-dimensional measures obtained from laser-digitized, virtual specimens. As shown in a comparison with a principal-component analysis of the virtual specimens, the artificial neural network approach provided more biologically meaningful information than the conventional multivariate analysis approach. The methodology discovered partitions and hierarchical clusters consistent with anthropoid systematics, from the species (or subspecies) level to the highest categories, by sorting and allocating upper and lower molar teeth. As one might expect, measures of upper molars were richer in phenetic information than those of lower molars, even among the anatomically diverse platyrrhines. We also show that reducing taxonomic noise (i.e. biological variation) by limiting the analysis to a monophyletic subset improves discrimination.

Algorithms↗
PubMed1997

Efficient Partition of Learning Data Sets for Neural Network Training.

This study investigates the emerging possibilities of combining unsupervised and supervised learning in neural network ensembles. Such strategy is used to get an efficient partition of a noisy input data set in order to focus the training of neural networks on the most complex and informative domains of the data set and accelerate the learning phase. The proposed algorithm provides a good prediction accuracy using fewer cases from non-informative domains according to a correlative measure of dependency between cases of the training set. This measure takes into account internal relationships amid analyzed data and can be used to cluster neighbor cases in a multidimensional space and to filter out the outliers. The possible relation of the proposed algorithm to brain processing occurring in the thalamo-cortical pathway is discussed.

Journal Article↗
PubMed1999

Analysis of gene expression data using self-organizing maps.

DNA microarray technologies together with rapidly increasing genomic sequence information is leading to an explosion in available gene expression data. Currently there is a great need for efficient methods to analyze and visualize these massive data sets. A self-organizing map (SOM) is an unsupervised neural network learning algorithm which has been successfully used for the analysis and organization of large data files. We have here applied the SOM algorithm to analyze published data of yeast gene expression and show that SOM is an excellent tool for the analysis and visualization of gene expression profiles.

Algorithms↗
PubMed2006

A modular analysis of breast cancer reveals a novel low-grade molecular signature in estrogen receptor-positive tumors.

PURPOSE: Previous reports using genome-wide gene expression data to classify breast tumors have typically used standard unsupervised or supervised techniques, both of which have known limitations. We hypothesized that novel clinically relevant information could be revealed in these data sets by an alternative analytic approach. Using a recently described algorithm, signature analysis (SA), we identified "modules," comprising groups of tightly coexpressed genes that are conditionally linked to particular tumors, in a series of breast tumor gene expression profiles. EXPERIMENTAL DESIGN AND RESULTS: The SA successfully identified multiple breast cancer modules specifically linked to distinct biological functions. We identified a novel module, TuM1, whose presence was not readily discernible by conventional clustering techniques. The TuM1 module is expressed in a subset of estrogen receptor (ER)-positive tumors and is significantly enriched with genes involved in apoptosis and cell death. Clinically, TuM1-expressing tumors are associated with low histopathologic grade, and this association is independent of the inherent ER status of a tumor. We confirmed the robustness and general applicability of TuM1 module by demonstrating its association with low tumor grade in multiple independent breast cancer data sets generated using different array technologies. In vitro, the TuM1 module is down-regulated in ER+ MCF7 cells upon treatment with tamoxifen, suggesting that TuM1 expression may be dependent on active signaling by ER. Initial data is also suggestive that TuM1 expression may be clinically associated with a patient's response to antihormonal therapy. CONCLUSION: Our results suggest that modular-based approaches toward gene expression data can prove useful in identifying novel, robust, and biologically relevant signatures even from data sets that have been the subject of substantial prior analysis.

Breast Neoplasms↗
PubMed2006

Semi-supervised discovery of differential genes.

BACKGROUND: Various statistical scores have been proposed for evaluating the significance of genes that may exhibit differential expression between two or more controlled conditions. However, in many clinical studies to detect clinical marker genes for example, the conditions have not necessarily been controlled well, thus condition labels are sometimes hard to obtain due to physical, financial, and time costs. In such a situation, we can consider an unsupervised case where labels are not available or a semi-supervised case where labels are available for a part of the whole sample set, rather than a well-studied supervised case where all samples have their labels. RESULTS: We assume a latent variable model for the expression of active genes and apply the optimal discovery procedure (ODP) proposed by Storey (2005) to the model. Our latent variable model allows gene significance scores to be applied to unsupervised and semi-supervised cases. The ODP framework improves detectability by sharing the estimated parameters of null and alternative models of multiple tests over multiple genes. A theoretical consideration leads to two different interpretations of the latent variable, i.e., it only implicitly affects the alternative model through the model parameters, or it is explicitly included in the alternative model, so that the interpretations correspond to two different implementations of ODP. By comparing the two implementations through experiments with simulation data, we have found that sharing the latent variable estimation is effective for increasing the detectability of truly active genes. We also show that the unsupervised and semi-supervised rating of genes, which takes into account the samples without condition labels, can improve detection of active genes in real gene discovery problems. CONCLUSION: The experimental results indicate that the ODP framework is effective for hypotheses including latent variables and is further improved by sharing the estimations of hidden variables over multiple tests.

Algorithms↗
PubMed2006

Novel unsupervised feature filtering of biological data.

MOTIVATION: Many methods have been developed for selecting small informative feature subsets in large noisy data. However, unsupervised methods are scarce. Examples are using the variance of data collected for each feature, or the projection of the feature on the first principal component. We propose a novel unsupervised criterion, based on SVD-entropy, selecting a feature according to its contribution to the entropy (CE) calculated on a leave-one-out basis. This can be implemented in four ways: simple ranking according to CE values (SR); forward selection by accumulating features according to which set produces highest entropy (FS1); forward selection by accumulating features through the choice of the best CE out of the remaining ones (FS2); backward elimination (BE) of features with the lowest CE. RESULTS: We apply our methods to different benchmarks. In each case we evaluate the success of clustering the data in the selected feature spaces, by measuring Jaccard scores with respect to known classifications. We demonstrate that feature filtering according to CE outperforms the variance method and gene-shaving. There are cases where the analysis, based on a small set of selected features, outperforms the best score reported when all information was used. Our method calls for an optimal size of the relevant feature set. This turns out to be just a few percents of the number of genes in the two Leukemia datasets that we have analyzed. Moreover, the most favored selected genes turn out to have significant GO enrichment in relevant cellular processes.

Artificial Intelligence↗