Single Linkage Clustering Research Articles

Novel protein domains and motifs in the marine planctomycete Rhodopirellula baltica

The planctomycetes are a phylum of bacteria that have a unique cell compartmentalisation and yeast-like budding cell division and peptidoglycan-less proteinaceous cell walls. We wished to further our understanding of these unique organisms at the molecular level by searching for conserved amino acid sequence motifs and domains in the proteins encoded by Rhodopirellula baltica. Using BLAST and single-linkage clustering, we have discovered several new protein domains and sequence motifs in this planctomycete. R. baltica has multiple members of the newly discovered GEFGR protein family and the ASPIC C-terminal domain family, whilst most other organisms for which whole genome sequence is available have no more than one. Many of the domains and motifs appear to be restricted to the planctomycetes. It is possible that these protein domains and motifs may have been lost or replaced in other phyla, or they may have undergone multiple duplication events in the planctomycete lineage. One of the novel motifs probably represents a novel N-terminal export signal peptide. With their unique cell biology, it may be that the planctomycete cell compartmentalisation plan in particular needs special membrane transport mechanisms. The discovery of these new domains and motifs, many of which are associated with secretion and cell-surface functions, will help to stimulate experimental work and thus enhance further understanding of this fascinating group of organisms.

A hybrid clustering approach to recognition of protein families in 114 microbial genomes

BackgroundGrouping proteins into sequence-based clusters is a fundamental step in many bioinformatic analyses (e.g., homology-based prediction of structure or function). Standard clustering methods such as single-linkage clustering capture a history of cluster topologies as a function of threshold, but in practice their usefulness is limited because unrelated sequences join clusters before biologically meaningful families are fully constituted, e.g. as the result of matches to so-called promiscuous domains. Use of the Markov Cluster algorithm avoids this non-specificity, but does not preserve topological or threshold information about protein families.ResultsWe describe a hybrid approach to sequence-based clustering of proteins that combines the advantages of standard and Markov clustering. We have implemented this hybrid approach over a relational database environment, and describe its application to clustering a large subset of PDB, and to 328577 proteins from 114 fully sequenced microbial genomes. To demonstrate utility with difficult problems, we show that hybrid clustering allows us to constitute the paralogous family of ATP synthase F1 rotary motor subunits into a single, biologically interpretable hierarchical grouping that was not accessible using either single-linkage or Markov clustering alone. We describe validation of this method by hybrid clustering of PDB and mapping SCOP families and domains onto the resulting clusters.ConclusionHybrid (Markov followed by single-linkage) clustering combines the advantages of the Markov Cluster algorithm (avoidance of non-specific clusters resulting from matches to promiscuous domains) and single-linkage clustering (preservation of topological information as a function of threshold). Within the individual Markov clusters, single-linkage clustering is a more-precise instrument, discerning sub-clusters of biological relevance. Our hybrid approach thus provides a computationally efficient approach to the automated recognition of protein families for phylogenomic analysis.

Multi-attribute scoring method for mobile telecommunication subscribers

The measurement of customer value has become a key issue for developing and maintaining successful customer relationship. To measure the value of each customer, it is necessary to consider not only the direct monetary term but also the potential customer value. In this paper, we propose a multi-attribute customer evaluation procedure for mobile telecommunication subscribers. At first, customers are clustered using a single linkage clustering method, then each segment is evaluated based on data envelopment analysis (DEA), which is frequently applied for multi-attribute evaluation problem. Finally, neural network model is used to evaluate each individual customer's value based on DEA results. The proposed method is applied to the mobile telecommunication subscriber data. It is expected that telecommunication companies can focus on target marketing to maximize the overall profit based on the efficiency score of customers.

Single-Linkage Clustering for Optimal Classification in Piecewise Affine Regression

When performing regression with piecewise affine maps, the most challenging task is to classify the data points, i.e. to correctly attribute a data point to the affine submodel that most likely generated it. In this paper, we consider a regression scheme similar to the one proposed in (Ferrari-Trecate et al., 2001; Ferrari-Trecate et al., 2003) that reduces the classification step to a clustering problem in presence of outliers. However instead of the K-means procedure adopted in (Ferrari-Trecate et al., 2001; Ferrari-Trecate et al., 2003), we propose the use of single-linkage clustering that estimates automatically the number of submodels composing the piecewise affine map. Moreover we prove that, under mild assunlptions on the data set, single-linkage clustering can guarantee optimal classification in presence of bounded noise.

Formation of machine groups and part families: a modified SLC method and comparative study

This paper introduces a modified single linkage clustering heuristic (MOD‐SC). The proposed MOD‐SLC objective is to test the application of Baroni‐Urban and Buser (BUB) similarity coefficient to the manufacturing cell formation (MCF) problem instead of Jaccard’s similarity coefficient. The MOD‐SLC has been compared and evaluated against three cluster formation‐based heuristics for MCF. The three heuristics are: the single linkage clustering, enhanced rank order clustering, and direct clustering algorithm. The MCF methods considered in this comparative and evaluative study belong to the cluster formation approach of solving the MCF problem. The comparison and evaluation are performed using four published performance measures. A total of 25 published and ten hypothetical and randomly generated problem data sets are used in the proposed evaluative study. Results analysis is carried out to test and validate the proposed BUB based MOD‐SLC. Finally the pros and cons of each method are stated and discussed.

Genome-Wide Functional Classification/Identification of Prokaryotic Transmembrane Proteins Based on Transmembrane Topology Similarity

Functions of transmembrane (TM) proteins holding 20-30% [1] in individual proteomes are difficult to be inferred comprehensively based on only sequence similarity, which is partly because TM protein sequences of known function are lacking. Recent studies, however, revealed that TM protein functions are closely correlated to their TM topologies, i.e., the number of TM segments (TMSs), positions of TMSs and the N-tail location [5]. In this study, we propose a new method for the comprehensive classification/identification of TM protein functions by single-linkage clustering based on TM topology similarity. Applying this method to TM protein sequences predicted from 87 sequenced prokaryotic genomes, it was realized that TM topology information improved functional classification/identification greatly.

Automated de novo identification of repeat sequence families in sequenced genomes.

Repetitive sequences make up a major part of eukaryotic genomes. We have developed an approach for the de novo identification and classification of repeat sequence families that is based on extensions to the usual approach of single linkage clustering of local pairwise alignments between genomic sequences. Our extensions use multiple alignment information to define the boundaries of individual copies of the repeats and to distinguish homologous but distinct repeat element families. When tested on the human genome, our approach was able to properly identify and group known transposable elements. The program, should be useful for first-pass automatic classification of repeats in newly sequenced genomes.

Density Visualization and Mode Hunting Using Trees

Mode hunting, a task related to cluster analysis and density estimation, is carried out in a variety of fields. The objective is to find (distinct) regions of the feature space that have relatively high density. Although existing algorithms for clustering and density estimation can be applied to this problem, they have difficulties dealing with datasets that have large numbers of observations and/or variables. This article presents and discusses a mode-hunting algorithm, based on recursively partitioning the feature space, that is particularly applicable to such data. The algorithm is fast, but avoids many of the pitfalls of other fast methods (e.g., single link clustering). The results from the analysis are amenable to visualization as a binary tree, thus allowing us to gain further insights about the data. We apply the method to a number of simulated and real datasets.

Knowledge-based selection of targets for structural genomics.

The problem of rational target selection for protein structure determination in structural genomics projects on microbes is addressed. A flexible computational procedure is described that directly incorporates the whole body of annotation available in the PEDANT genome database into the sequence clustering and selection process in order to identify proteins that are likely to possess currently unknown structural domains. Filtering out gene products based on predicted structural features, such as known three-dimensional structures and transmembrane regions, allows one to reduce the complexity of neighbor relationships between sequences and all but eliminates the need for further partitioning of single-linkage clusters into disjoint protein groups corresponding to homologous families. The results of a large-scale computation experiment in which exemplary target selection for 32 prokaryotic genomes was conducted are presented.

A graph-based clustering method for a large set of sequences using a graph partitioning algorithm.

A graph-based clustering method is proposed to cluster protein sequences into families, which automatically improves clusters of the conventional single linkage clustering method. Our approach formulates sequence clustering problem as a kind of graph partitioning problem in a weighted linkage graph, which vertices correspond to sequences, edges correspond to higher similarities than given threshold and are weighted by their similarities. The effectiveness of our method is shown in comparison with InterPro families in all mouse proteins in SWISS-PROT. The result clusters match to InterPro families much better than the single linkage clustering method. 77% of proteins in InterPro families are classified into appropriate clusters.

GeneRAGE: a robust algorithm for sequence clustering and domain detection.

Efficient, accurate and automatic clustering of large protein sequence datasets, such as complete proteomes, into families, according to sequence similarity. Detection and correction of false positive and negative relationships with subsequent detection and resolution of multi-domain proteins. A new algorithm for the automatic clustering of protein sequence datasets has been developed. This algorithm represents all similarity relationships within the dataset in a binary matrix. Removal of false positives is achieved through subsequent symmetrification of the matrix using a Smith-Waterman dynamic programming alignment algorithm. Detection of multi-domain protein families and further false positive relationships within the symmetrical matrix is achieved through iterative processing of matrix elements with successive rounds of Smith-Waterman dynamic programming alignments. Recursive single-linkage clustering of the corrected matrix allows efficient and accurate family representation for each protein in the dataset. Initial clusters containing multi-domain families, are split into their constituent clusters using the information obtained by the multi-domain detection step. This algorithm can hence quickly and accurately cluster large protein datasets into families. Problems due to the presence of multi-domain proteins are minimized, allowing more precise clustering information to be obtained automatically. GeneRAGE (version 1.0) executable binaries for most platforms may be obtained from the authors on request. The system is available to academic users free of charge under license.

Microbial food web structure in the Arabian Sea: a US JGOFS study

One of the main objectives of the Joint Global Ocean Flux Studies (JGOFS) program is to develop an understanding of the factors controlling organic carbon production in the ocean and the time-varying vertical flux of carbon from surface waters (US JGOFS (1990) US JGOFS Planning Report Number 11; Sarmiento and Armstrong (1997) US JGOFS Synthesis and Modeling Project Implementation Plan). A considerable amount of evidence suggests that carbon cycling and the potential for exporting carbon from ocean systems is a function of food web structure. As part of the US JGOFS Arabian Sea Studies, the biomass of planktonic organisms, ranging from heterotrophic bacteria through microplankton-sized organisms, was estimated using a variety of methods including flow cytometry and microscopy. This is a first attempt to combine biomass data from a number of sources, evaluate the structure of the food web, examine changes in food web structure in relation to seasonal or spatial features of the study area, and look for indications of how changing structure affects carbon-cycling processes. Biomass in the upper 100 m of the water column ranged from approximately 1.5 to >5.2 gC m −2. Heterotrophic bacteria (Hbac) made up from 16 and 44% of the biomass; autotrophs comprised 43–64%; and the remainder was made up of nano- and microheterotrophs. Autotrophs and nano- and microheterotrophs showed a general pattern of higher values at coastal stations, with the lowest values offshore. Heterotrophic bacteria (Hbac) showed no significant spatial variations. The Spring Intermonsoon and early NE Monsoon were dominated by autotrophic picoplankton, Prochlorococcus and Synechococcus. The late NE Monsoon and late SW Monsoon periods showed an increase in the larger size fractions of the primary producers. At several stations during the SW Monsoon, autotrophic microplankton, primarily diatoms and Phaeocystis colonies, predominated. Increases in the size of autotrophs were also reflected in increasing sizes of nano- and microheterorophs. The biomass estimates based on cytometry and microscopy are consistent with measurement of pigments, POC and PON. Changes in community structure were assessed using the percent similarity index (PSI) in conjunction with multidimensional scaling (MDS) or single-linkage clustering analysis to show how assemblages differed among cruises and stations. Station clustering reflected environmental heterogeneity, and many of the conspicuous changes could be associated with changes in temperature, salinity and nutrient concentrations. Despite inherent problems in combining data from a variety of sources, the present community biomass estimates were well constrained by bulk measurements such as Chl a, POC and PON, and by comparisons with other quantitative and qualitative studies. The most striking correlation between food web structure and carbon cycling was the dominance of large phytoplankton, primarily diatoms, and the seasonal maxima of mass flux during the SW Monsoon. High nutrient conditions associated with upwelling during the SW Monsoon would explain the predominance of diatoms during this season. The sinking of large, ungrazed diatom cells is one possible explanation for the flux observations, but may not be consistent with the observation of concurrent increases in larger microzooplankton consumers (heterotrophic dinoflagellates and ciliates) and mesozooplankton during this season. Food-web structure during the early NE Monsoon and Intermonsoons suggests carbon cycling by the microbial community predominated.

Hierarchical Cluster Analysis Applied to Workers Exposures in Fiberglass Insulation Manufacturing

The objectives of this study were to explore the application of cluster analysis to the characterization of multiple exposures in industrial hygiene practice and to compare exposure groupings based on the result from cluster analysis with that based on non-measurement-based approaches commonly used in epidemiology. Cluster analysis was performed for 37 workers simultaneously exposed to three agents (endotoxin, phenolic compounds and formaldehyde) in fiberglass insulation manufacturing. Different clustering algorithms, including complete-linkage (or farthest-neighbor), single-linkage (or nearest-neighbor), group-average and model-based clustering approaches, were used to construct the tree structures from which clusters can be formed. Differences were observed between the exposure clusters constructed by these different clustering algorithms. When contrasting the exposure classification based on tree structures with that based on non-measurement-based information, the results indicate that the exposure clusters identified from the tree structures had little in common with the classification results from either the traditional exposure zone or the work group classification approach. In terms of the defining homogeneous exposure groups or from the standpoint of health risk, some toxicological normalization in the components of the exposure vector appears to be required in order to form meaningful exposure groupings from cluster analysis. Finally, it remains important to see if the lack of correspondence between exposure groups based on epidemiological classification and measurement data is a peculiarity of the data or a more general problem in multivariate exposure analysis.

Hierarchical Cluster Analysis Applied to Workers Exposures in Fiberglass Insulation Manufacturing

The objectives of this study were to explore the application of cluster analysis to the characterization of multiple exposures in industrial hygiene practice and to compare exposure groupings based on the result from cluster analysis with that based on non-measurement-based approaches commonly used in epidemiology. Cluster analysis was performed for 37 workers simultaneously exposed to three agents (endotoxin, phenolic compounds and formaldehyde) in fiberglass insulation manufacturing. Different clustering algorithms, including complete-linkage (or farthest-neighbor), single-linkage (or nearest-neighbor), group-average and model-based clustering approaches, were used to construct the tree structures from which clusters can be formed. Differences were observed between the exposure clusters constructed by these different clustering algorithms. When contrasting the exposure classification based on tree structures with that based on non-measurement-based information, the results indicate that the exposure clusters identified from the tree structures had little in common with the classification results from either the traditional exposure zone or the work group classification approach. In terms of the defining homogeneous exposure groups or from the standpoint of health risk, some toxicological normalization in the components of the exposure vector appears to be required in order to form meaningful exposure groupings from cluster analysis. Finally, it remains important to see if the lack of correspondence between exposure groups based on epidemiological classification and measurement data is a peculiarity of the data or a more general problem in multivariate exposure analysis.

Phylogenetic analysis of the third hsp70 homolog in Escherichia coli; a novel member of the Hsc66 subfamily and its possible co-chaperone.

Novel members of the highly conserved protein family, Hsp70, have been found in the complete sequences of several genomes. To elucidate a phylogenetic relationship among Hsp70 proteins of Escherichia coli, we searched all open reading frames derived from 13 complete genomes for Hsp70/actin-related proteins by the single-linkage clustering method. Phylogenetic analysis of this superfamily revealed that E. coli possesses at least three Hsp70 homologs (DnaK, Hsc66 and Hsc62). We found that Hsc62, which is the product of hscC, is a new member of the Hsc66 subfamily, and is specific to E. coli. The analysis also suggested that YegD of E. coli is closely related to the actin family, which consists of the actin, FtsA and MreB subfamilies. A further database search revealed that two dnaJ homologs, ybeS and ybeV, were located on the opposite strand near hscC. Consequently, E. coli seems to have three gene clusters composed of DnaK and DnaJ homologs.

Application of global optimization to particle identification using light scattering

Numerical methods of solving the inverse light scattering problem for spheres are presented. The methods are based on two stochastic global optimization techniques: Deep's random search and the multilevel single-linkage clustering analysis due to Rinnooy Kan and Timmer. Computational examples show that the radius and the refractive index of spheres comparable with or larger than the wavelength of light can be recovered from multiangle scattering data. While the random search approach is faster, the clustering analysis is shown to be more reliable. A general discussion of the clustering method is also given.

DIVCLUS: an automatic method in the GEANFAMMER package that finds homologous domains in single- and multi-domain proteins.

Large-scale determination of relationships between the proteins produced by genome sequences is now common. All protein sequences are matched and those that have high match scores are clustered into families. In cases where the proteins are built of several domains or duplication modules, this can lead to misleading results. Consider the very simple example of three proteins: 1, formed by duplication modules A and B; 2, formed by duplication modules B' and C; and 3, formed by duplication modules C' and D. Duplication modules B and B' are homologous, as are C and C'. Matching the sequences of 1, 2 and 3 followed by simple single-linkage clustering would put all three in the same family, even though proteins 1 and 3 are not related. This is because the different parts of 2 match 1 and 3. This paper describes a procedure, DIVCLUS, that divides such complex clusters of partially related sequences into simple clusters that contain only related duplication modules. In the example just given, it would produce two groups of sequences: the first with domains B of sequence 1 and B of sequence 2, and the second with domain C of sequence 2 and C of sequence 3. DIVCLUS is part of a package called GEANFAMMER, for GEnome ANalysis and protein FAMily MakER. The package automates the detection of families of duplication modules from a protein sequence database. DIVCLUS has been applied to the division of single-linkage clusters generated from the protein sequences of six completely sequenced bacterial genomes. Out of 12 013 genes in these six genomes, 4563 single- and multi-domain sequences formed 1071 complex clusters. Application of the DIVCLUS program resolved these clusters into 2113 clusters corresponding to single duplication modules. The perl5 program and its documentation are available at the following address: http://www.mrc-lmb.cam.ac.uk/genomes/ and by anonymous ftp at ftp.mrc-lmb.cam.ac.uk in the directory /pub/genomes/Software/. sat@mrc-lmb.cam.ac.uk; jong@mrc-lmb. cam.ac.uk

Copper(II) six-membered rings: a molecular mechanics and cluster analysis

An agglomerative, hierarchical single-link clustering method was used to separate the different conformations of all the saturated six-membered copper(II) rings found in the Cambridge Structure Database. Three percent of all rings were found to be in the sofa and twist-boat conformations, one percent were boats and the rest chairs. There was no correlation between the coordination geometry of the copper(II) ion and the conformation of the six-membered ring. Changing the ligating atoms from nitrogen to sulfur had no effect on the conformational distribution of the six-membered rings. Molecular mechanics calculations were used to establish the reasons why high energy non-chair conformations wre adopted in certain cases.

Monotone linkage clustering and quasi-concave set functions

Greedily seriating objects one by one is implicitly employed in many heuristic clustering procedures which can be described in terms of a linkage function measuring entity-to-set dissimilarities. A well-known clustering technique, single linkage clustering, can be considered as an example of the seriation procedures (based actually on the minimum spanning tree construction) leading to the global maximum of a corresponding “minimum split” set function. The purpose of this work is to extend this property to the wide class of so-called monotone linkages. It is shown that the minimum split functions of monotone linkages can be maximized greedily. Moreover, this class of set functions is proven to coincide with the class of so-called quasi-concave set functions.

Part family formation for variety reduction in flexible manufacturing systems

Presents the similarity coefficient method for group technology to alleviate the part family formation problem in flexible manufacturing systems (FMS). Part families are formed in FMS in order to take advantage of part similarities in design and manufacture. Parts coding and classification analysis (PCA) has constituted the bulk of part family formation techniques in practice. Using shape‐based features for grouping is very labour intensive at the coding and classification stages. As a means of alleviating the latter problem, presents an approach for converting the weighted codes of the PCA to similarity coefficient measures. Uses a clustering algorithm to identify the part families. Presents a numerical example that compares the single and average linkage clustering technologies. An experimental investigation of the two methods showed that the average linkage clustering (ALC) performs better than the single linkage clustering (SLC) technology in minimizing intercellular materials handling costs.