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 379 records · Page 21Linked to original sources
PubMed2001

Characterization of the geographical origin of Pecorino Sardo cheese by casein stable isotope ((13)c/(12)c and (15)n/(14)n) ratios and free amino acid ratios.

The stable isotope ratios ((13)C/(12)C and (15)N/(14)N) of casein measured by isotope ratio mass spectrometry (IRMS) and some free amino acid ratios (His/Pro, Ile/Pro, Met/Pro, and Thr/Pro) determined by HPLC in samples of ewes' milk cheese from Sardinia, Sicily, and Apulia were found to be parameters independent of ripening time. Multivariate data treatments performed by applying both unsupervised (principal component analysis and cluster analysis) and supervised [linear discriminant analysis (LDA)] methods revealed good discrimination possibilities for the cheeses according to place of origin. In this respect, particularly significant were the variables Ile/Pro, Thr/Pro, (13)C/(12)C, and (15) N/(14)N ratios on which basis 100% discrimination and classification of the samples by LDA was obtained.

Amino Acids↗
PubMed2003

Genome-wide comparison of human keratinocyte and squamous cell carcinoma responses to UVB irradiation: implications for skin and epithelial cancer.

To gain insight into the transformation of epidermal cells into squamous carcinoma cells (SCC), we compared the response to ultraviolet B radiation (UVB) of normal human epidermal keratinocytes (NHEK) versus their transformed counterpart, SCC, using biological and molecular profiling. DNA microarray analyses (Affymetrix), approximately 12000 genes) indicated that the major group of upregulated genes in keratinocytes fall into three categories: (i). antiapoptotic and cell survival factors, including chemokines of the CXC/CC subfamilies (e.g. IL-8, GRO-1, -2, -3, SCYA20), growth factors (e.g. HB-EGF, CTGF, INSL-4), and proinflammatory mediators (e.g. COX-2, S100A9), (ii). DNA repair-related genes (e.g. GADD45, ERCC, BTG-1, Histones), and (iii). ECM proteases (MMP-1, -10). The major downregulated genes are DeltaNp63 and PUMILIO, two potential markers for the maintenance of keratinocyte stem cells. NHEK were found to be more resistant than SCC to UVB-induced apoptosis and this resistance was mainly because of the protection from cell death by secreted survival factors, since it can be transferred from NHEK to SCC cultures by the conditioned medium. Whereas the response of keratinocytes to UVB involved regulation of key checkpoint genes (p53, MDM2, p21(Cip1), DeltaNp63), as well as antiapoptotic and DNA repair-related genes - no or little regulation of these genes was observed in SCC. The effect of UVB on NHEK and SCC resulted in upregulation of 251 and 127 genes, respectively, and downregulation of 322 genes in NHEK and 117 genes in SCC. To further analyse these changes, we used a novel unsupervised coupled two-way clustering method that allowed the identification of groups of genes that clearly partitioned keratinocytes from SCC, including a group of genes whose constitutive expression levels were similar before UVB. This allowed the identification of discriminating genes not otherwise revealed by simple static comparison in the absence of UVB irradiation. The implication of the changes in gene profile in keratinocytes for epithelial cancer is discussed.

Apoptosis↗
PubMed2004

Different gene expression profiles between microsatellite instability-high and microsatellite stable colorectal carcinomas.

Recent molecular genetic studies have revealed that two major types of genomic instabilities, chromosomal instability (CIN) and microsatellite instability (MSI), exist in colorectal carcinomas. In order to clarify the molecular signature related to the CIN and MSI in colorectal carcinomas, we performed transcriptomic expression analysis on eight microsatellite instability-high (MSI-H) colorectal carcinomas and compared the results obtained with that of nine microsatellite stable (MSS) colorectal carcinomas using oligonucleotide microarrays containing 17 334 known genes and 1331 unknown genes or expression sequence tags (ESTs). Unsupervised two-way hierarchical clustering with 5724 genes successfully classified tumors from normal mucosa, and displayed a distinctive MSI-H carcinomas subgroup. Based on intensive filtering, 57 known genes and eight ESTs were found to be highly relevant to the differentiation of MSI-H and MSS colorectal carcinomas. These genes successfully distinguish the new test set of six MSI-H and five MSS colorectal carcinomas. Many up- and downregulated genes in MSI-H colorectal carcinomas were related to the previously reported phenotypic characteristics; increased mucin production and intense peritumoral immune response in MSI-H carcinomas. Some of these differences were confirmed by semiquantitative reverse transcription-PCR and immunohistochemical analysis. Our findings indicate that there are many different genetic and transcriptomic characteristics between MSI-H and MSS colorectal carcinomas, and some of these differently expressed genes can be used as diagnostic or prognostic markers.

Base Sequence↗
PubMed2003

Predicting HIV drug resistance with neural networks.

MOTIVATION: Drug resistance is a very important factor influencing the failure of current HIV therapies. The ability to predict the drug resistance of HIV protease mutants may be useful in developing more effective and longer lasting treatment regimens. METHODS: The HIV resistance is predicted to two current protease inhibitors, Indinavir and Saquinavir. The problem was approached from two perspectives. First, a predictor was constructed based on the structural features of the HIV protease-drug inhibitor complex. A particular structure was represented by its list of contacts between the inhibitor and the protease. Next, a classifier was constructed based on the sequence data of various drug resistant mutants. In both cases, self-organizing maps were first used to extract the important features and cluster the patterns in an unsupervised manner. This was followed by subsequent labelling based on the known patterns in the training set. RESULTS: The prediction performance of the classifiers was measured by cross-validation. The classifier using the structure information correctly classified previously unseen mutants with an accuracy of between 60 and 70%. Several architectures were tested on the more abundant sequence data. The best single classifier provided an accuracy of 68% and a coverage of 69%. Multiple networks were then combined into various majority voting schemes. The best combination yielded an average of 85% coverage and 78% accuracy on previously unseen data. This is more than two times better than the 33% accuracy expected from a random classifier.

Algorithms↗
PubMed2005

Oligonucleotide microarray analysis of distinct gene expression patterns in colorectal cancer tissues harboring BRAF and K-ras mutations.

Various types of human cancers harbor BRAF somatic mutations, leading researchers to seek molecular targets for BRAF inhibitors. A mutually exclusive relationship has been observed between the BRAF-V600E mutation and K-ras mutations, suggesting that the BRAF-V600E mutation may differ from the other BRAF mutant types. Here, we used microarray analysis to examine differences between the BRAF and K-ras mutant colorectal samples and within the BRAF group (V600E versus non-V600E), in the hope that the identified gene sets could form the basis for new target development. Eleven colorectal cancers (CRCs) with BRAF mutations and nine with K-ras mutations were examined by high-density microarray analysis. We also tested whether other significant genetic or clinical status involved in CRC development, such as APC and TP53 mutations, MSI and TNM-Duke's staging, were related with the observed BRAF- or K-ras associated expression profiles. Unsupervised two-way hierarchical clustering and multidimensional scaling revealed that the differentially expressed genes clustered according to the mutation status of BRAF and K-ras, and that samples with the BRAF-V600E and non-V600E mutants could be distinguished from each other by gene profiling. Examination of TNM-Duke's staging, MSI and mutations in APC and TP53 revealed that these significant mutations could not account for the hierarchical clustering results observed in our study. We herein identified distinct gene expression patterns and gene sets that may form the basis for identification of BRAF-targeting molecules or provide researchers with a better understanding of the molecular pathogenesis underlying RAS-RAF signaling.

Colorectal Neoplasms↗
PubMed2006

Profiling markers of prognosis in colorectal cancer.

PURPOSE: Colorectal cancer is one of the most common forms of cancer in developed nations and the incidence of this disease is increasing. There is a need to further stratify prognostically distinct groups of colorectal cancer, and the purpose of this study was to identify prognostically significant immunohistochemical marker profiles in colorectal cancer. EXPERIMENTAL DESIGN: In this study, a range (n = 23) of markers [pRb, p16, p21, p27, p53, proliferating cell nuclear antigen, cyclin D1, bcl-2, epidermal growth factor receptor, C-erb-B2, topoisomerase-I, liver fatty acid-binding protein, matrix metalloproteinases (MMP) 1-3, 7, 9, and 13, MT1-MMP, MT2-MMP, and tissue inhibitors of MMP 1-3] of putative prognostic significance have been investigated by immunohistochemistry on formalin-fixed, wax-embedded sections in a series (n = 90) of stage III (Dukes C) colorectal cancers. An immunohistochemical score based on the intensity of immunoreactivity and, where relevant, the proportion of immunoreactive cells was established for each marker. RESULTS: Unsupervised two-dimensional hierarchical cluster analysis identified three distinct cluster groups (designated groups 1-3) with different marker profiles. There were significant survival differences between groups 1 and 2 (log rank = 11.48; P = 0.0007) and between groups 1 and 3 (log rank = 8.32; P = 0.0039). Multivariate analysis showed that the complete marker profile was independently the most significant prognostic factor (hazard ratio, 2.27; 95% confidence interval, 1.15-4.48; P = 0.004). CONCLUSIONS: This study has identified an immunohistochemical marker profile of colorectal cancer and showed that it is an independent indicator of prognosis in this type of cancer.

Adult↗
PubMed2006

Wnt-1 is dominant over neu in specifying mammary tumor expression profiles.

Wnt-1 and Neu collaborate to induce mammary tumors in bitransgenic mice carrying both MMTV-Wnt-1 and MMTV-Neu. In this report, gene expression profiles were determined for tumors from these bitransgenic mice, and compared with expression profiles of tumors from mice singly transgenic for MMTV-Wnt-1 or MMTV-Neu. While very different from tumors arising in MMTV-Neu transgenic mice, tumors from these bitransgenic mice were found not to have identifiable differences from tumors from MMTV-Wnt-1 transgenic mice, using clustering and multidimensional scaling analyses (unsupervised and supervised), One-way Analysis of Variance (ANOVA), and two sample t test (the later two of which were combined with false discovery rate computation). These observations suggest that Wnt-1 is dominant over Neu in specifying mammary tumor expression profiles.

Animals↗
PubMed2003

Gene expression profiles obtained from fine-needle aspirations of breast cancer reliably identify routine prognostic markers and reveal large-scale molecular differences between estrogen-negative and estrogen-positive tumors.

PURPOSE: The purpose of this study was to determine whether comprehensive transcriptional profiles (TPs) can be obtained from single-passage fine-needle aspirations (FNAs) of breast cancer and to explore whether profiles capture routine clinicopathological parameters. EXPERIMENTAL DESIGN: Expression profiles were available on 38 patients with stage I-III breast cancer who underwent FNA at the time of diagnosis. [(33)P]dCTP-labeled cDNA probes were generated and hybridized to cDNA membrane microarrays that contained 30,000 human sequence clones, including 10,890 expressed sequence tags. RESULTS: The median total RNA yield from the biopsies was 2 micro g (range, 1-25 micro g). The cellular composition of each biopsy was examined and, on average, 79% of the cells were cancer cells. TP was successfully performed on all 38 of the biopsies. Unsupervised complete-linkage hierarchical clustering with all of the biopsies revealed an association between estrogen receptor (ER) status and gene expression profiles. There was a strong correlation between ER status determined by TP and measured by routine immunohistochemistry (P = 0.001). A similar strong correlation was seen with HER-2 status determined by fluorescent in situ hybridization (P = 0.0002). Using the first 18 cases as the discovery set, we established a cutoff of 2.0 and 18.0 for ER and HER-2 mRNA levels, respectively, to distinguish clinically-negative from -positive cases. We also identified 105 genes (excluding the ER gene) the expression of which correlated highly with clinical ER status. Twenty tumors were used for prospective validation. HER-2 status was correctly identified in all 20 of the cases, based on HER-2 mRNA content detected by TP. ER status was correctly identified in 19 of 20 cases. When the marker set of 105 genes was used to prospectively predict ER status, TP-based classification correctly identified 9 of 10 of the ER-positive and 7 of 10 of the ER-negative tumors. We also explored supervised cluster analysis using various functional categories of genes, and we observed a clear separation between ER-negative and ER-positive tumors when genes involved in signal transduction were used for clustering. CONCLUSIONS: These results demonstrate that comprehensive TP can be performed on FNA biopsies. TPs reliably measure conventional single-gene prognostic markers such as ER and HER-2. A complex pattern of genes (not including ER) can also be used to predict clinical ER status. These results demonstrate that needle biopsy-based diagnostic microarray tests may be developed that could capture conventional prognostic information but may also contain additional clinical information that cannot currently be measured with other methods.

Adult↗
PubMed1991

Application of pattern recognition methods to the analysis and classification of toxicological data derived from proton nuclear magnetic resonance spectroscopy of urine.

A computer-based pattern recognition (PR) approach has been applied to the classification and interrogation of 1H NMR-generated urinalysis data, in a variety of experimental toxicity states in the rat. 1H NMR signal intensities for each endogenous urinary metabolite were regarded as coordinates in multidimensional space and analyzed using PR methods, through which the dimensionality was reduced for display and categorization purposes. The changes in the NMR spectral patterns were characterized by 17 metabolic dimensions, which were then analyzed by employing the unsupervised learning methods of hierarchical cluster analysis, two-dimensional nonlinear map (NLM) analysis, and two- or three-dimensional principal components analysis (PCA). Different types of toxin (hepatotoxins and cortical and papillary nephrotoxins) were classified according to NMR-detectable biochemical effects. PCA provided consistently better results than NLMs in terms of discrimination of toxicity type, and maps based on correlation matrices also gave improved discrimination over those based on raw data. Various refinements in the data analysis were investigated, including taking NMR urinalysis data at three time points after exposure of the rats to six different nephrotoxins, as well as employing a dual-scoring system (time and magnitude of change). The maps generated from the time-course information produced the best discrimination between nephrotoxins from different classes. The robustness of the classification methods (in particular NLMs and PCA based on correlation matrices) and the influence of the addition of new scored biochemical data, reflecting dose-response situations, nutritional effects on toxicity, sex differences in biochemical response to toxins, the addition of a new toxin class (cadmium chloride, a testicular toxin and renal carbonic anhydrase inhibitor), and an additional metabolite descriptor (creatine), to the PR analysis were also evaluated. Initial training set maps were fundamentally stable to the addition of new data, and both NLM and PCA methods correctly "predicted" the toxicological effects from NMR data for test compounds, suggesting that the approach using PR and 1H NMR urinalysis for the generation and classification of acute toxicological data has wide applicability.

Animals↗
PubMed2005

An improved cluster labeling method for support vector clustering.

The support vector clustering (SVC) algorithm is a recently emerged unsupervised learning method inspired by support vector machines. One key step involved in the SVC algorithm is the cluster assignment of each data point. A new cluster labeling method for SVC is developed based on some invariant topological properties of a trained kernel radius function. Benchmark results show that the proposed method outperforms previously reported labeling techniques.

Algorithms↗
PubMed2005

Clustering ensembles: models of consensus and weak partitions.

Clustering ensembles have emerged as a powerful method for improving both the robustness as well as the stability of unsupervised classification solutions. However, finding a consensus clustering from multiple partitions is a difficult problem that can be approached from graph-based, combinatorial, or statistical perspectives. This study extends previous research on clustering ensembles in several respects. First, we introduce a unified representation for multiple clusterings and formulate the corresponding categorical clustering problem. Second, we propose a probabilistic model of consensus using a finite mixture of multinomial distributions in a space of clusterings. A combined partition is found as a solution to the corresponding maximum-likelihood problem using the EM algorithm. Third, we define a new consensus function that is related to the classical intraclass variance criterion using the generalized mutual information definition. Finally, we demonstrate the efficacy of combining partitions generated by weak clustering algorithms that use data projections and random data splits. A simple explanatory model is offered for the behavior of combinations of such weak clustering components. Combination accuracy is analyzed as a function of several parameters that control the power and resolution of component partitions as well as the number of partitions. We also analyze clustering ensembles with incomplete information and the effect of missing cluster labels on the quality of overall consensus. Experimental results demonstrate the effectiveness of the proposed methods on several real-world data sets.

Algorithms↗
PubMed1998

A fuzzy vessel tracking algorithm for retinal images based on fuzzy clustering.

In this paper we present a new unsupervised fuzzy algorithm for vessel tracking that is applied to the detection of the ocular fundus vessels. The proposed method overcomes the problems of initialization and vessel profile modeling that are encountered in the literature and automatically tracks fundus vessels using linguistic descriptions like "vessel" and "nonvessel." The main tool for determining vessel and nonvessel regions along a vessel profile is the fuzzy C-means clustering algorithm that is fed with properly preprocessed data. Additional procedures for checking the validity of the detected vessels and handling junctions and forks are also presented. The application of the proposed algorithm to fundus images and simulated vessels resulted in very good overall performance and consistent estimation of vessel parameters.

Algorithms↗
PubMed2001

Cluster analysis of comparative genomic hybridization (CGH) data using self-organizing maps: application to prostate carcinomas.

Comparative genomic hybridization (CGH) is a modern genetic method which enables a genome-wide survey of chromosomal imbalances. For each chromosome region, one obtains the information whether there is a loss or gain of genetic material, or whether there is no change at that region. Usually it is not possible to evaluate all 46 chromosomes of a metaphase, therefore several (up to 20 or more) metaphases are analyzed per individual, and expressed as average. Mostly one does not study one individual alone but groups of 20-30 individuals. Therefore, large amounts of data quickly accumulate which must be put into a logical order. In this paper we present the application of a self-organizing map (Genecluster) as a tool for cluster analysis of data from pT2N0 prostate cancer cases studied by CGH. Self-organizing maps are artificial neural networks with the capability to form clusters on the basis of an unsupervised learning rule, i.e., in our examples it gets the CGH data as only information (no clinical data). We studied a group of 40 recent cases without follow-up, an older group of 20 cases with follow-up, and the data set obtained by pooling both groups. In all groups good clusterings were found in the sense that clinically similar cases were placed into the same clusters on the basis of the genetic information only. The data indicate that losses on chromosome arms 6q, 8p and 13q are all frequent in pT2N0 prostatic cancer, but the loss on 8p has probably the largest prognostic importance.

Carcinoma↗
PubMed2002

Chromosomal regions in prostatic carcinomas studied by comparative genomic hybridization, hierarchical cluster analysis and self-organizing feature maps.

Comparative genomic hybridization (CGH) is an established genetic method which enables a genome-wide survey of chromosomal imbalances. For each chromosome region, one obtains the information whether there is a loss or gain of genetic material, or whether there is no change at that place. Therefore, large amounts of data quickly accumulate which must be put into a logical order. Cluster analysis can be used to assign individual cases (samples) to different clusters of cases, which are similar and where each cluster may be related to a different tumour biology. Another approach consists in a clustering of chromosomal regions by rewriting the original data matrix, where the cases are written as rows and the chromosomal regions as columns, in a transposed form. In this paper we applied hierarchical cluster analysis as well as two implementations of self-organizing feature maps as classical and neuronal tools for cluster analysis of CGH data from prostatic carcinomas to such transposed data sets. Self-organizing maps are artificial neural networks with the capability to form clusters on the basis of an unsupervised learning rule. We studied a group of 48 cases of incidental carcinomas, a tumour category which has not been evaluated by CGH before. In addition we studied a group of 50 cases of pT2N0-tumours and a group of 20 pT3N0-carcinomas. The results show in all case groups three clusters of chromosomal regions, which are (i) normal or minimally affected by losses and gains, (ii) regions with many losses and few gains and (iii) regions with many gains and few losses. Moreover, for the pT2N0- and pT3N0-groups, it could be shown that the regions 6q, 8p and 13q lay all on the same cluster (associated with losses), and that the regions 9q and 20q belonged to the same cluster (associated with gains). For the incidental cancers such clear correlations could not be demonstrated.

Chromosome Mapping↗
PubMed2004

Clustering of diverse genomic data using information fusion.

MOTIVATION: Genome sequencing projects and high-through-put technologies like DNA and Protein arrays have resulted in a very large amount of information-rich data. Microarray experimental data are a valuable, but limited source for inferring gene regulation mechanisms on a genomic scale. Additional information such as promoter sequences of genes/DNA binding motifs, gene ontologies, and location data, when combined with gene expression analysis can increase the statistical significance of the finding. This paper introduces a machine learning approach to information fusion for combining heterogeneous genomic data. The algorithm uses an unsupervised joint learning mechanism that identifies clusters of genes using the combined data. RESULTS: The correlation between gene expression time-series patterns obtained from different experimental conditions and the presence of several distinct and repeated motifs in their upstream sequences is examined here using publicly available yeast cell-cycle data. The results show that the combined learning approach taken here identifies correlated genes effectively. The algorithm provides an automated clustering method, but allows the user to specify apriori the influence of each data type on the final clustering using probabilities. AVAILABILITY: Software code is available by request from the first author. CONTACT: jkasturi@cse.psu.edu.

Algorithms↗
PubMed2005

Clustering algorithms and other exploratory methods for microarray data analysis.

OBJECTIVES: We introduce methods for the exploratory analysis of microarray data, especially focusing on cluster algorithms. Benefits and problems are discussed. METHODS: We describe application and suitability of unsupervised learning methods for the classification of gene expression data. Cluster algorithms are treated in more detail, including assessment of cluster quality. RESULTS: When dealing with microarray data, most cluster algorithms must be applied with caution. As long as the structure of the true generating models of such data is not fully understood, the use of simple algorithms seems to be more appropriate than the application of complex black-box algorithms. New methods explicitly targeted to the analysis of microarray data are increasingly being developed in order to increase the amount of useful information extracted from the experiments. CONCLUSIONS: Unsupervised methods can be a helpful tool for the analysis of microarray data, but a critical choice of the algorithm and a careful interpretation of the results are required in order to avoid false conclusions.

Algorithms↗
PubMed2000

Reclassification as supervised clustering.

In some branches of science, such as molecular biology, classes may be defined but not completely trusted. Sometimes posterior analysis proves them to be partially incorrect. Despite its relevance, this phenomenon has not received much attention within the neural computation community. We define reclassification as the task of redefining some given classes by maximum likelihood learning in a model that contains both supervised and unsupervised information. This approach leads to supervised clustering with an additional complexity penalizing term on the number of new classes. As a proof of concept, a simple reclassification algorithm is designed and applied to a data set of gene sequences. To test the performance of the algorithm, two of the original classes are merged. The algorithm is capable of unraveling the original three-class hidden structure, in contrast to the unsupervised version (K-means); moreover, it predicts the subdivision of one of the original classes into two different ones.

Algorithms↗
PubMed2005

Signature of B-CLL with different prognosis by Shrunken centroids of surface antigen expression profiling.

With the aim of identifying the immunophenotypic profile of B-cell chronic lymphocytic leukemia (B-CLL) subsets with different prognosis, we investigated by flow cytometry the expression of 36 surface antigens in 123 cases, all with survivals. By analyzing results with unsupervised (hierarchical and K-means clustering) algorithms, three distinct immunophenotypic groups (I, II, and III) were identified, group I (51/123) with longer survivals, as compared to the group II (36/123) and III (36/123). The immunophenotypic signatures of these groups, as determined by applying the nearest Shrunken centroids method as class predictor, were characterized by the coordinated and differential expression of 12 surface markers, that is, group I: above-average expression of CD62L, CD54, CD49c, and CD25, below-average expression of CD38; group II: above-average expression of CD38, CD49d, CD29, and CD49e; and group III: below-average expression of the above markers, overexpression of CD23, CD20, SmIg, and CD79b. As opposed to groups II-III, group I B-CLLs lacked expression of ZAP-70 and activation-induced cytidine deaminase in the majority of cases, while more frequently had mutated IgV(H) genes and IgV(H) mutations consistent with antigen-driven selection. Our findings contribute to improve the immunophenotypical identification of disease subsets with different prognosis and suggest a set of surface antigens to be employed as prognosticators in routine diagnostic/prognostic procedures.

Adult↗