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 613 records · Page 34Linked to original sources
PubMed1995

Machine learning to support diagnostics in the domain of asymptomatic liver disease.

Machine learning procedures, in unsupervised and supervised manner, can enable their users to achieve knowledge hardly comprehensible by even the best experts. This is true also if the clinical knowledge has been carefully assembled in a prospective way. A data set including 165 patients with elevated routine laboratory tests was extensively studied according to clinical history, laboratory profile and liver biopsy. Unsupervised learning by Kohonen feature map disclosed 4 groups of patients: the largest one with no or slight histopathological changes (116) and three smaller, more homogenous, with more diseased patients. Standardized histopathological scorings of the liver specimens defined patients into two groups. Fifty-eight of them were, according to the analysis, recommended for a liver biopsy, due to more severe degrees of inflammation and fibrosis. One-hundred and seven of the patients, in whom liver biopsy was retrospectively considered unnecessary, had only minor degrees of inflammation, fibrosis and/or steatosis. Supervised learning, using the inductive systems based on Quinlan's ID3 and CART algorithms, extracted knowledge in the form of decision trees. This approach could define a need for biopsy either with a very few significant findings or by pathways containing quotients and multiplications of the different basic items. These procedures were analyzed and compared for their theoretical and applicative performances. The cluster and Fischerian discriminant analyses were performed in order to compare the classification performance. The medical appropriateness of the obtained results is satisfying, therefore decision support systems, outlined in this study, should be evaluated in wider clinical practice. To achieve this goal, an example of a Medical Logical Module (MLM), based on the Arden Syntax, is given.

Adult↗
PubMed2005

Employing ICA and SOM for spike sorting of multielectrode recordings from CNS.

For classification of action potential shapes in multineuron recordings, we present a spike sorting system employing independent component analysis (ICA) and an unsupervised artificial neural network (Kohonen's self-organizing map, SOM). We focus on how ICA in the first stage of the spike sorting system can be used to address specific problems arising in recordings using multielectrode arrays in the CNS. Using real data recorded from the pontine nuclei in rats and simulated data, we evaluate the performance of several ICA algorithms to remove cross-talk between electrodes using data from continuous recording (or simulation). When using cut-out data, the standard format of extracellular spike recordings, new problems emerge and robust algorithms are needed. We demonstrate that several ICA algorithms show a good performance on cut-out data from multielectrode array recordings (simulated and real data). In tetrode recordings the same neuron is purposely recorded by several electrodes simultaneously and we show, how independent component analysis can be used in this case to identify redundant information and hence to compress relevant information, improving subsequent clustering of a SOM.

Action Potentials↗
PubMed2007

Feature subset selection and ranking for data dimensionality reduction.

A new unsupervised forward orthogonal search (FOS) algorithm is introduced for feature selection and ranking. In the new algorithm, features are selected in a stepwise way, one at a time, by estimating the capability of each specified candidate feature subset to represent the overall features in the measurement space. A squared correlation function is employed as the criterion to measure the dependency between features and this makes the new algorithm easy to implement. The forward orthogonalization strategy, which combines good effectiveness with high efficiency, enables the new algorithm to produce efficient feature subsets with a clear physical interpretation.

Algorithms↗
PubMed2006

Functional interaction of nitrogenous organic bases with cytochrome P450: a critical assessment and update of substrate features and predicted key active-site elements steering the access, binding, and orientation of amines.

The widespread use of nitrogenous organic bases as environmental chemicals, food additives, and clinically important drugs necessitates precise knowledge about the molecular principles governing biotransformation of this category of substrates. In this regard, analysis of the topological background of complex formation between amines and P450s, acting as major catalysts in C- and N-oxidative attack, is of paramount importance. Thus, progress in collaborative investigations, combining physico-chemical techniques with chemical-modification as well as genetic engineering experiments, enables substantiation of hypothetical work resulting from the design of pharmacophores or homology modelling of P450s. Based on a general, CYP2D6-related construct, the majority of prospective amine-docking residues was found to cluster near the distal heme face in the six known SRSs, made up by the highly variant helices B', F and G as well as the N-terminal portion of helix C and certain beta-structures. Most of the contact sites examined show a frequency of conservation < 20%, hinting at the requirement of some degree of conformational versatility, while a limited number of amino acids exhibiting a higher level of conservation reside close to the heme core. Some key determinants may have a dual role in amine binding and/or maintenance of protein integrity. Importantly, a series of non-SRS elements are likely to be operative via long-range effects. While hydrophobic mechanisms appear to dominate orientation of the nitrogenous compounds toward the iron-oxene species, polar residues seem to foster binding events through H-bonding or salt-bridge formation. Careful uncovering of structure-function relationships in amine-enzyme association together with recently developed unsupervised machine learning approaches will be helpful in both tailoring of novel amine-type drugs and early elimination of potentially toxic or mutagenic candidates. Also, chimeragenesis might serve in the construction of more efficient P450s for activation of amine drugs and/or bioremediation.

Amines↗
PubMed2005

Identifying critical variables of principal components for unsupervised feature selection.

Principal components analysis (PCA) is probably the best-known approach to unsupervised dimensionality reduction. However, axes of the lower-dimensional space, ie., principal components (PCs), are a set of new variables carrying no clear physical meanings. Thus, interpretation of results obtained in the lower-dimensional PCA space and data acquisition for test samples still involve all of the original measurements. To deal with this problem, we develop two algorithms to link the physically meaningless PCs back to a subset of original measurements. The main idea of the algorithms is to evaluate and select feature subsets based on their capacities to reproduce sample projections on principal axes. The strength of the new algorithms is that the computaion complexity involved is significantly reduced, compared with the data structural similarity-based feature evaluation.

Algorithms↗
PubMed2005

Surface-antigen expression profiling (SEP) in B-cell chronic lymphocytic leukemia (B-CLL): Identification of markers with prognostic relevance.

Studies of gene expression profiling (GEP) have been successfully used for the identification of molecules to be employed as potential prognosticators. With the aim of identifying the immunophenotypic profile of B-CLL subsets with different prognoses, we investigated by flow cytometry the expression of 36 surface antigens in 117 cases, 113 with survival data. In analogy with GEP, results were analyzed by applying unsupervised hierarchical algorithms (surface-antigen expression profiling, SEP). Distinct immunophenotypic groups (A, B1, B2 and C) were identified, group C (57/117) with longer survivals, as compared to groups A (23/117), B1 (16/117) and B2 (21/117). The immunophenotypic signatures of these groups were characterized by the coordinated and differential over-expression of: i) CD62L, CD54 and CD49c (group C); ii) CD38 and CD49d (group A); iii) none of the above markers (group B1 and B2). Other molecules were either not expressed, widely expressed by all samples, or were variably expressed within the observed B-CLL subgroups, although without a clearly distinguishable pattern. By employing an identical approach for investigating the reactivity of B-cell panel monoclonal antibodies (B-mAbs) in B-CLLs (29 cases) and in 19 B and non-B leukemia/lymphoma cell lines, we found mAbs (B012, B001, B006, B018, B019, B020, B017) mainly unreactive in all the samples, mAbs (B002, B010, B013, B014, B015) strongly reactive in B-CLLs and B-cell lines but not in non-B-cell lines, and mAbs recognizing antigens variably expressed in cell lines and B-CLLs. A hierarchical clustering focused on B-CLLs alone, combining reactivity values for B-mAbs with the expression of CD62L and CD38, these latter antigens identified as leader markers of B-CLL subsets with different prognosis, demonstrated a correlation between CD62L expression and the reactivity of B007, B003, B011 and B005 mAbs. These mAbs may represent potentially novel markers with prognostic relevance in B-CLLs.

ADP-ribosyl Cyclase 1↗
PubMed2006

Unsupervised contour closure algorithm for range image edge-based segmentation.

This paper presents an efficient technique for extracting closed contours from range images' edge points. Edge points are assumed to be given as input to the algorithm (i.e., previously computed by an edge-based range image segmentation technique). The proposed approach consists of three steps. Initially, a partially connected graph is generated from those input points. Then, the minimum spanning tree of that graph is computed. Finally, a postprocessing technique generates a single path through the regions' boundaries by removing noisy links and closing open contours. The novelty of the proposed approach lies in the fact that, by representing edge points as nodes of a partially connected graph, it reduces the contour closure problem to a minimum spanning tree partitioning problem plus a cost function minimization stage to generate closed contours. Experimental results with synthetic and real range images, together with comparisons with a previous technique, are presented.

Algorithms↗
PubMed2005

Analysis of differentially-regulated genes within a regulatory network by GPS genome navigation.

MOTIVATION: A critical challenge of the post-genomic era is to understand how genes are differentially regulated even when they belong to a given network. Because the fundamental mechanism controlling gene expression operates at the level of transcription initiation, computational techniques have been developed that identify cis regulatory features and map such features into expression patterns to classify genes into distinct networks. However, these methods are not focused on distinguishing between differentially regulated genes within a given network. Here we describe an unsupervised machine learning method, termed GPS for gene promoter scan, that discriminates among co-regulated promoters by simultaneously considering both cis-acting regulatory features and gene expression. GPS is particularly useful for knowledge discovery in environments with reduced datasets and high levels of uncertainty. RESULTS: Application of this method to the enteric bacteria Escherichia coli and Salmonella enterica uncovered novel members, as well as regulatory interactions in the regulon controlled by the PhoP protein that were not discovered using previous approaches. The predictions made by GPS were experimentally validated to establish that the PhoP protein uses multiple mechanisms to control gene transcription, and is a central element in a highly connected network. AVAILABILITY: The scripts and programs used in this work are accessible from the gps-tools.wustl.edu website. Data and predictions are available by request.

Algorithms↗
PubMed2006

Super-resolution registration using tissue-classified distance fields.

We present a method for registering the position and orientation of bones across multiple computed-tomography (CT) volumes of the same subject. The method is subvoxel accurate, can operate on multiple bones within a set of volumes, and registers bones that have features commensurate in size to the voxel dimension. First, a geometric object model is extracted from a reference volume image. We use then unsupervised tissue classification to generate from each volume to be registered a super-resolution distance field--a scalar field that specifies, at each point, the signed distance from the point to a material boundary. The distance fields and the geometric bone model are finally used to register an object through the sequence of CT images. In the case of multiobject structures, we infer a motion-directed hierarchy from the distance-field information that allows us to register objects that are not within each other's capture region. We describe a validation framework and evaluate the new technique in contrast with grey-value registration. Results on human wrist data show average accuracy improvements of 74% over grey-value registration. The method is of interest to any intrasubject, same-modality registration applications where subvoxel accuracy is desired.

Algorithms↗
PubMed2006

Mycobacterium tuberculosis complex genetic diversity: mining the fourth international spoligotyping database (SpolDB4) for classification, population genetics and epidemiology.

BACKGROUND: The Direct Repeat locus of the Mycobacterium tuberculosis complex (MTC) is a member of the CRISPR (Clustered regularly interspaced short palindromic repeats) sequences family. Spoligotyping is the widely used PCR-based reverse-hybridization blotting technique that assays the genetic diversity of this locus and is useful both for clinical laboratory, molecular epidemiology, evolutionary and population genetics. It is easy, robust, cheap, and produces highly diverse portable numerical results, as the result of the combination of (1) Unique Events Polymorphism (UEP) (2) Insertion-Sequence-mediated genetic recombination. Genetic convergence, although rare, was also previously demonstrated. Three previous international spoligotype databases had partly revealed the global and local geographical structures of MTC bacilli populations, however, there was a need for the release of a new, more representative and extended, international spoligotyping database. RESULTS: The fourth international spoligotyping database, SpolDB4, describes 1939 shared-types (STs) representative of a total of 39,295 strains from 122 countries, which are tentatively classified into 62 clades/lineages using a mixed expert-based and bioinformatical approach. The SpolDB4 update adds 26 new potentially phylogeographically-specific MTC genotype families. It provides a clearer picture of the current MTC genomes diversity as well as on the relationships between the genetic attributes investigated (spoligotypes) and the infra-species classification and evolutionary history of the species. Indeed, an independent Naïve-Bayes mixture-model analysis has validated main of the previous supervised SpolDB3 classification results, confirming the usefulness of both supervised and unsupervised models as an approach to understand MTC population structure. Updated results on the epidemiological status of spoligotypes, as well as genetic prevalence maps on six main lineages are also shown. Our results suggests the existence of fine geographical genetic clines within MTC populations, that could mirror the passed and present Homo sapiens sapiens demographical and mycobacterial co-evolutionary history whose structure could be further reconstructed and modelled, thereby providing a large-scale conceptual framework of the global TB Epidemiologic Network. CONCLUSION: Our results broaden the knowledge of the global phylogeography of the MTC complex. SpolDB4 should be a very useful tool to better define the identity of a given MTC clinical isolate, and to better analyze the links between its current spreading and previous evolutionary history. The building and mining of extended MTC polymorphic genetic databases is in progress.

Computational Biology↗
PubMed2006

Self-organizing neural networks to support the discovery of DNA-binding motifs.

Identification of the short DNA sequence motifs that serve as binding targets for transcription factors is an important challenge in bioinformatics. Unsupervised techniques from the statistical learning theory literature have often been applied to motif discovery, but effective solutions for large genomic datasets have yet to be found. We present here three self-organizing neural networks that have applicability to the motif-finding problem. The core system in this study is a previously described SOM-based motif-finder named SOMBRERO. The motif-finder is integrated in this work with a SOM-based method that automatically constructs generalized models for structurally related motifs and initializes SOMBRERO with relevant biological knowledge. A self-organizing tree method that displays the relationships between various motifs is also presented, and it is shown that such a method can act as an effective structural classifier of novel motifs. The performance of the three self-organizing neural networks is evaluated here using various datasets.

Algorithms↗
PubMed2004

An unsupervised approach to identify molecular phenotypic components influencing breast cancer features.

To discover a biological basis for clinical subgroupings within breast cancers, we applied principal components (PCs) analysis to cDNA microarray data from 36 breast cancers. We correlated the resulting PCs with clinical features. The 35 PCs discovered were ranked in order of their impact on gene expression patterns. Interestingly, PC 7 identified a unique subgroup consisting of estrogen receptor (ER); (+) African-American patients. This group exhibited global molecular phenotypes significantly different from both ER (-) African-American women and ER (+) or ER (-) Caucasian women (P < 0.001). Additional significant PCs included PC 4, correlating with lymph node metastasis (P = 0.04), and PC 10, with tumor stage (stage 2 versus stage 3; P = 0.007). These results provide a molecular phenotypic basis for the existence of a biologically unique subgroup comprising ER (+) breast cancers from African-American patients. Moreover, these findings illustrate the potential of PCs analysis to detect molecular phenotypic bases for relevant clinical or biological features of human tumors in general.

Adult↗
PubMedDate not supplied

Essential latent knowledge for protein-protein interactions: analysis by an unsupervised learning approach.

Protein-protein interactions play a number of central roles in many cellular functions, including DNA replication, transcription and translation, signal transduction, and metabolic pathways. A recent increase in the number of protein-protein interactions has made predicting unknown protein-protein interactions important for the understanding of living cells. However, the protein-protein interactions experimentally obtained so far are often incomplete and contradictory and, consequently, existing computational prediction methods have integrated evidence (latent knowledge of proteins) from different and more reliable sources. Analyzing the relationships between proteins and the latent knowledge is important to understanding the cellular processes. For this analysis, we propose a new probabilistic model for protein-protein interactions by considering the latent knowledge of proteins. We further present an efficient learning algorithm for this model, based on an EM algorithm. Experimental results have shown that in a supervised test setting, the proposed method outperformed five other competing methods by a statistically significant factor in all cases. Using the probability parameters of a trained model, we have further shown the latent knowledge that is essential to predicting protein-protein interactions. Overall, our experimental results confirm that our proposed model is especially effective for analyzing protein-protein interactions from a viewpoint of the latent knowledge of proteins.

Algorithms↗
PubMed2006

MUSA: a parameter free algorithm for the identification of biologically significant motifs.

MOTIVATION: The ability to identify complex motifs, i.e. non-contiguous nucleotide sequences, is a key feature of modern motif finders. Addressing this problem is extremely important, not only because these motifs can accurately model biological phenomena but because its extraction is highly dependent upon the appropriate selection of numerous search parameters. Currently available combinatorial algorithms have proved to be highly efficient in exhaustively enumerating motifs (including complex motifs), which fulfill certain extraction criteria. However, one major problem with these methods is the large number of parameters that need to be specified. RESULTS: We propose a new algorithm, MUSA (Motif finding using an UnSupervised Approach), that can be used either to autonomously find over-represented complex motifs or to estimate search parameters for modern motif finders. This method relies on a biclustering algorithm that operates on a matrix of co-occurrences of small motifs. The performance of this method is independent of the composite structure of the motifs being sought, making few assumptions about their characteristics. The MUSA algorithm was applied to two datasets involving the bacterium Pseudomonas putida KT2440. The first one was composed of 70 sigma(54)-dependent promoter sequences and the second dataset included 54 promoter sequences of up-regulated genes in response to phenol, as suggested by quantitative proteomics. The results obtained indicate that this approach is very effective at identifying complex motifs of biological significance. AVAILABILITY: The MUSA algorithm is available upon request from the authors, and will be made available via a Web based interface.

Algorithms↗
PubMed2005

A 3-d surface reconstruction approach based on postnonlinear ICA model.

Photometric stereo technique deals with the reconstruction of three-dimensional (3-D) shape of an object by using several images of the same surface taken from the same viewpoint but under illuminations from different directions. In this paper, we propose a new photometric stereo scheme based on a new reflectance model and the postnonlinear (PNL) independent components analysis (ICA) method. The proposed nonlinear reflectance model consists of diffuse components and specular components for modeling the surface reflectance of a stereo object in an image. Unlike the previous approaches, these two components are not separated and processed individually in the proposed model. An unsupervised learning adaptation algorithm is developed to estimate the reflectance model based on image intensities. In this algorithm, the PNL ICA method is used to obtain the surface normal on each point of an image. Then, the 3-D surface model is reconstructed based on the estimated surface normal on each point of image by using the enforcing integrability method. Two experiments are performed to assess the performance of the proposed approach. We test our algorithm on synthetically generated images for the reconstruction of surface of objects and on a number of real images captured from the Yale Face Database B. These testing images contain variability due to illumination and varying albedo in each point of surface of human faces. All the experimental results are compared to those of the existing photometric stereo approaches tested on the same images. The results clearly indicate the superiority of the proposed nonlinear reflectance model over the conventional Lambertian model and the other linear hybrid reflectance model.

Algorithms↗
PubMed2006

Marked genomic differences characterize primary and secondary glioblastoma subtypes and identify two distinct molecular and clinical secondary glioblastoma entities.

Glioblastoma is classified into two subtypes on the basis of clinical history: "primary glioblastoma" arising de novo without detectable antecedent disease and "secondary glioblastoma" evolving from a low-grade astrocytoma. Despite their distinctive clinical courses, they arrive at an indistinguishable clinical and pathologic end point highlighted by widespread invasion and resistance to therapy and, as such, are managed clinically as if they are one disease entity. Because the life history of a cancer cell is often reflected in the pattern of genomic alterations, we sought to determine whether primary and secondary glioblastomas evolve through similar or different molecular pathogenetic routes. Clinically annotated primary and secondary glioblastoma samples were subjected to high-resolution copy number analysis using oligonucleotide-based array comparative genomic hybridization. Unsupervised classification using genomic nonnegative matrix factorization methods identified three distinct genomic subclasses. Whereas one corresponded to clinically defined primary glioblastomas, the remaining two stratified secondary glioblastoma into two genetically distinct cohorts. Thus, this global genomic analysis showed wide-scale differences between primary and secondary glioblastomas that were previously unappreciated, and has shown for the first time that secondary glioblastoma is heterogeneous in its molecular pathogenesis. Consistent with these findings, analysis of regional recurrent copy number alterations revealed many more events unique to these subclasses than shared. The pathobiological significance of these shared and subtype-specific copy number alterations is reinforced by their frequent occurrence, resident genes with clear links to cancer, recurrence in diverse cancer types, and apparent association with clinical outcome. We conclude that glioblastoma is composed of at least three distinct molecular subtypes, including novel subgroups of secondary glioblastoma, which may benefit from different therapeutic strategies.

Astrocytoma↗
PubMed2004

Identification of cell cycle-regulated genes in fission yeast.

Cell cycle progression is both regulated and accompanied by periodic changes in the expression levels of a large number of genes. To investigate cell cycle-regulated transcriptional programs in the fission yeast Schizosaccharomyces pombe, we developed a whole-genome oligonucleotide-based DNA microarray. Microarray analysis of both wild-type and cdc25 mutant cell cultures was performed to identify transcripts whose levels oscillated during the cell cycle. Using an unsupervised algorithm, we identified 747 genes that met the criteria for cell cycle-regulated expression. Peaks of gene expression were found to be distributed throughout the entire cell cycle. Furthermore, we found that four promoter motifs exhibited strong association with cell cycle phase-specific expression. Examination of the regulation of MCB motif-containing genes through the perturbation of DNA synthesis control/MCB-binding factor (DSC/MBF)-mediated transcription in arrested synchronous cdc10 mutant cell cultures revealed a subset of functional targets of the DSC/MBF transcription factor complex, as well as certain gene promoter requirements. Finally, we compared our data with those for the budding yeast Saccharomyces cerevisiae and found approximately 140 genes that are cell cycle regulated in both yeasts, suggesting that these genes may play an evolutionarily conserved role in regulation of cell cycle-specific processes. Our complete data sets are available at http://giscompute.gis.a-star.edu.sg/~gisljh/CDC.

Algorithms↗
PubMed2005

Clear cell sarcoma of the kidney: up-regulation of neural markers with activation of the sonic hedgehog and Akt pathways.

PURPOSE AND EXPERIMENTAL DESIGN: Clear cell sarcoma of the kidney (CCSK), the second most common renal tumor in children, poses significant diagnostic challenges. No positive diagnostic markers are available, and the pathogenesis of CCSK remains an enigma. To address these challenges, the gene expression patterns of 14 CCSKs were compared with 15 Wilms tumors and 3 fetal kidney samples using oligonucleotide arrays. RESULTS: Using unsupervised methods, the gene expression profile of CCSK was distinctive: differentially expressed genes could largely be grouped into four categories: (a) a wide variety of neural markers, (b) members of the Sonic hedgehog pathway, (c) members of the phosphoinositide-3-kinase/Akt cell proliferation pathway, and (d) known therapeutic targets. Corresponding changes in critical proteins using Western blot and/or immunohistochemistry confirmed the up-regulation of these pathways and proteins. In particular, CD117 and epidermal growth factor receptor are up-regulated at the protein level in many CCSKs, providing potential therapeutic targets. One of the neural markers, nerve growth factor receptor, represents a promising diagnostic tool for CCSK. CONCLUSIONS: This study suggests that CCSKs arise within a renal mesenchymal cell that shows a wide variety of neural markers. As such, it seems to be susceptible to genetic changes also seen in a variety of other neuroectodermal and neuronal tumors, including activation of Sonic hedgehog and phosphoinositide-3-kinase/Akt pathways. Involvement of these pathways in CCSKs implicates their widening role in tumorigenesis.

Biomarkers, Tumor↗