A Mixture model with random-effects components for clustering correlated gene-expression profiles

Motivation: The clustering of gene profiles across some experimental conditions of interest contributes significantly to the elucidation of unknown gene function, the validation of gene discoveries and the interpretation of biological processes. However, this clustering problem is not straightforward as the profiles of the genes are not all independently distributed and the expression levels may have been obtained from an experimental design involving replicated arrays. Ignoring the dependence between the gene profiles and the structure of the replicated data can result in important sources of variability in the experiments being overlooked in the analysis, with the consequent possibility of misleading inferences being made. We propose a random-effects model that provides a unified approach to the clustering of genes with correlated expression levels measured in a wide variety of experimental situations. Our model is an extension of the normal mixture model to account for the correlations between the gene profiles and to enable covariate information to be incorporated into the clustering process. Hence the model is applicable to longitudinal studies with or without replication, for example, time-course experiments by using time as a covariate, and to cross-sectional experiments by using categorical covariates to represent the different experimental classes. Results: We show that our random-effects model can be fitted by maximum likelihood via the EM algorithm for which the E(expectation) and M(maximization) steps can be implemented in closed form. Hence our model can be fitted deterministically without the need for time-consuming Monte Carlo approximations. The effectiveness of our model-based procedure for the clustering of correlated gene profiles is demonstrated on three real datasets, representing typical microarray experimental designs, covering time-course, repeated-measurement and cross-sectional data. In these examples, relevant clusters of the genes are obtained, which are supported by existing gene-function annotation. A synthetic dataset is considered too.

Discovery of ultracompact dwarf galaxies in the Virgo Cluster

We have discovered nine ultracompact dwarf galaxies (UCDs) in the Virgo Cluster, extending samples of these objects outside the Fornax Cluster. Using the Two Degree Field (2dF) multifiber spectrograph on the Anglo-Australian Telescope, the new Virgo members were found among 1500 color-selected, starlike targets with 16: 0 < b(j) < 20.2 in a 2 degrees diameter field centered on M87 (NGC 4486). The newly found UCDs are comparable to the UCDs in the Fornax Cluster, with sizes less than or similar to 100 pc, -12.9 < M-B < -10.7, and exhibiting red absorption-line spectra, indicative of an older stellar population. The properties of these objects remain consistent with the tidal threshing model for the origin of UCDs from the surviving nuclei of nucleated dwarf elliptical galaxies disrupted in the cluster core but can also be explained as objects that were formed by mergers of star clusters created in galaxy interactions. The discovery that UCDs exist in Virgo shows that this galaxy type is probably a ubiquitous phenomenon in clusters of galaxies; coupled with their possible origin by tidal threshing, the UCD population is a potential indicator and probe of the formation history of a given cluster. We also describe one additional bright UCD with M-B = -12.0 in the core of the Fornax Cluster. We find no further UCDs in our Fornax Cluster Spectroscopic Survey down to bj 19.5 in two additional 2dF fields extending as far as 3 degrees from the center of the cluster. All six Fornax bright UCDs identified with 2dF lie within 0.degrees 5 (projected distance of 170 kpc) of the central elliptical galaxy NGC 1399.

On the statistical analysis of the GS-NS0 cell proteome: Imputation, clustering and variability testing

We have undertaken two-dimensional gel electrophoresis proteomic profiling on a series of cell lines with different recombinant antibody production rates. Due to the nature of gel-based experiments not all protein spots are detected across all samples in an experiment, and hence datasets are invariably incomplete. New approaches are therefore required for the analysis of such graduated datasets. We approached this problem in two ways. Firstly, we applied a missing value imputation technique to calculate missing data points. Secondly, we combined a singular value decomposition based hierarchical clustering with the expression variability test to identify protein spots whose expression correlates with increased antibody production. The results have shown that while imputation of missing data was a useful method to improve the statistical analysis of such data sets, this was of limited use in differentiating between the samples investigated, and highlighted a small number of candidate proteins for further investigation. (c) 2006 Elsevier B.V. All rights reserved.

The 2df SDSS LRG and QSO survey: evolution of the luminosity function of luminous red galaxies to z=0.6

We present new measurements of the luminosity function (LF) of luminous red galaxies (LRGs) from the Sloan Digital Sky Survey (SDSS) and the 2dF SDSS LRG and Quasar (2SLAQ) survey. We have carefully quantified, and corrected for, uncertainties in the K and evolutionary corrections, differences in the colour selection methods, and the effects of photometric errors, thus ensuring we are studying the same galaxy population in both surveys. Using a limited subset of 6326 SDSS LRGs (with 0.17 < z < 0.24) and 1725 2SLAQ LRGs (with 0.5 < z < 0.6), for which the matching colour selection is most reliable, we find no evidence for any additional evolution in the LRG LF, over this redshift range, beyond that expected from a simple passive evolution model. This lack of additional evolution is quantified using the comoving luminosity density of SDSS and 2SLAQ LRGs, brighter than M-0.2r - 5 log h(0.7) = - 22.5, which are 2.51 +/- 0.03 x 10(-7) L circle dot Mpc(-3) and 2.44 +/- 0.15 x 10(-7) L circle dot Mpc(-3), respectively (< 10 per cent uncertainty). We compare our LFs to the COMBO-17 data and find excellent agreement over the same redshift range. Together, these surveys show no evidence for additional evolution (beyond passive) in the LF of LRGs brighter than M-0.2r - 5 log h(0.7) = - 21 ( or brighter than similar to L-*).. We test our SDSS and 2SLAQ LFs against a simple 'dry merger' model for the evolution of massive red galaxies and find that at least half of the LRGs at z similar or equal to 0.2 must already have been well assembled (with more than half their stellar mass) by z similar or equal to 0.6. This limit is barely consistent with recent results from semi-analytical models of galaxy evolution.

The effect of local galaxy surface density on star formation for HI selected galaxies

We present the result of investigations into two theories to explain the star formation rate (SFR)-density relationship. For regions of high galaxy density, either there are fewer star-forming galaxies or galaxies capable of forming stars are present but some physical process is suppressing their star formation. We use H I Parkes All-Sky Survey's (HIPASS) HI detected galaxies and infrared and radio fluxes to investigate SFRs and efficiencies with respect to local surface density. For nearby (vel < 10 000 km s(-1)) H I galaxies, we find a strong correlation between H I mass and SFR. The number of H I galaxies decreases with increasing local surface density. For H I galaxies (1000 < vel < 6000 km s(-1)), there is no significant change in the SFR or the efficiency of star formation with respect to local surface density. We conclude that the SFR-density relation is due to a decrease in the number of H I star-forming galaxies in regions of high galaxy density and not to the suppression of star formation.

The survey for ionization in neutral gas galaxies. I. Description and initial results

We introduce the Survey for Ionization in Neutral Gas Galaxies (SINGG), a census of star formation in H I selected galaxies. The survey consists of H alpha and R-band imaging of a sample of 468 galaxies selected from the H I Parkes All Sky Survey (HIPASS). The sample spans three decades in H I mass and is free of many of the biases that affect other star-forming galaxy samples. We present the criteria for sample selection, list the entire sample, discuss our observational techniques, and describe the data reduction and calibration methods. This paper focuses on 93 SINGG targets whose observations have been fully reduced and analyzed to date. The majority of these show a single emission line galaxy (ELG). We see multiple ELGs in 13 fields, with up to four ELGs in a single field. All of the targets in this sample are detected in H alpha, indicating that dormant (non-star-forming) galaxies with M-H I greater than or similar to 3x10(7) M-circle dot are very rare. A database of the measured global properties of the ELGs is presented. The ELG sample spans 4 orders of magnitude in luminosity (H alpha and R band), and H alpha surface brightness, nearly 3 orders of magnitude in R surface brightness and nearly 2 orders of magnitude in H alpha equivalent width (EW). The surface brightness distribution of our sample is broader than that of the Sloan Digital Sky Survey (SDSS) spectroscopic sample, the EW distribution is broader than prism-selected samples, and the morphologies found include all common types of star-forming galaxies (e.g., irregular, spiral, blue compact dwarf, starbursts, merging and colliding systems, and even residual star formation in S0 and Sa spirals). Thus, SINGG presents a superior census of star formation in the local universe suitable for further studies ranging from the analysis of H II regions to determination of the local cosmic star formation rate density.

The survey for ionization in neutral gas galaxies. II. The star formation rate density of the local universe

We derive observed H alpha and R-band luminosity densities of an H I-selected sample of nearby galaxies using the SINGG sample to be l'(H alpha) = (9.4 +/- 1.8) x 10(38) h(70) ergs s(-1) Mpc(-3) for H alpha and l'(R) = (4.4 +/- 9.7) x 10(37) h(70) ergs s(-1) angstrom(-1) Mpc(-3) in the R band. This R-band luminosity density is approximately 70% of that found by the Sloan Digital Sky Survey. This leads to a local star formation rate density of log ((rho)over dot(SFR) [M-circle dot yr(-1) Mpc(-3)]) = -1.80(-0.07)(+0.13)(random) +/- 0.03(systematic) + log (h(70)) after applying a mean internal extinction correction of 0.82 mag. The gas cycling time of this sample is found to be t(gas) = 7.5(-2.1)(+1.3) Gyr, and the volume-averaged equivalent width of the SINGG galaxies is EW(H alpha) = 28.8(-4.7)(+7.2) angstrom (21.2-3.5+4.2 angstrom without internal dust correction). As with similar surveys, these results imply that (rho)over dot(SFR)(z) decreases drastically from z similar to 1.5 to the present. A comparison of the dynamical masses of the SINGG galaxies evaluated at their optical limits with their stellar and H I masses shows significant evidence of downsizing: the most massive galaxies have a larger fraction of their mass locked up in stars compared with H I, while the opposite is true for less massive galaxies. We show that the application of the Kennicutt star formation law to a galaxy having the median orbital time at the optical limit of this sample results in a star formation rate decay with cosmic time similar to that given by the. (rho)over dot(SFR)(z) evolution. This implies that the (rho)over dot(SFR)(z) evolution is primarily due to the secular evolution of galaxies, rather than interactions or mergers. This is consistent with the morphologies predominantly seen in the SINGG sample.

The 2dF-SDSS LRG and QSO survey: the star formation histories of luminous red galaxies

We present a detailed investigation into the recent star formation histories of 5697 luminous red galaxies (LRGs) based on the H delta (4101 angstrom), and [O II] (3727 angstrom) lines and the D4000 index. LRGs are luminous (L > 3L*) galaxies which have been selected to have photometric properties consistent with an old, passively evolving stellar population. For this study, we utilize LRGs from the recently completed 2dF-SDSS LRG and QSO Survey (2SLAQ). Equivalent widths of the H delta and [O II] lines are measured and used to define three spectral types, those with only strong H delta absorption (k+a), those with strong [O II] in emission (em) and those with both (em+a). All other LRGs are considered to have passive star formation histories. The vast majority of LRGs are found to be passive (similar to 80 per cent); however, significant numbers of k+a (2.7 per cent), em+a (1.2 per cent) and em LRGs (8.6 per cent) are identified. An investigation into the redshift dependence of the fractions is also performed. A sample of SDSS MAIN galaxies with colours and luminosities consistent with the 2SLAQ LRGs is selected to provide a low-redshift comparison. While the em and em+a fractions are consistent with the low-redshift SDSS sample, the fraction of k+a LRGs is found to increase significantly with redshift. This result is interpreted as an indication of an increasing amount of recent star formation activity in LRGs with redshift. By considering the expected lifetime of the k+a phase, the number of LRGs which will undergo a k+a phase can be estimated. A crude comparison of this estimate with the predictions from semi-analytic models of galaxy formation shows that the predicted level of k+a and em+a activities is not sufficient to reconcile the predicted mass growth for massive early types in a hierarchical merging scenario.

Symptom prevalence and clustering of symptoms in people living with chronic hepatitis C infection

Quality of life has been shown to be poor among people living with chronic hepatitis C However, it is not clear how this relates to the presence of symptoms and their severity. The aim of this study was to describe the typology of a broad array of symptoms that were attributed to hepatitis C virus (HCV) infection. Phase I used qualitative methods to identify symptoms. In Phase 2, 188 treatment-naive people living with HCV participated in a quantitative survey. The most prevalent symptom was physical tiredness (86%) followed by irritability (75%), depression (70%), mental tiredness (70%), and abdominal pain (68%). Temporal clustering of symptoms was reported in 62% of participants. Principal components analysis identified four symptom clusters: neuropsychiatric (mental tiredness, poor concentration, forgetfulness, depression, irritability, physical tiredness, and sleep problems); gastrointestinal (day sweats, nausea, food intolerance, night sweats, abdominal pain, poor appetite, and diarrhea); algesic (joint pain, muscle pain, and general body pain); and dysesthetic (noise sensitivity, light sensitivity, skin. problems, and headaches). These data demonstrate that symptoms are prevalent in treatment-naive people with HCV and support the hypothesis that symptom clustering occurs.

Efficient spatial clustering algorithm using binary tree

In this paper we present an efficient k-Means clustering algorithm for two dimensional data. The proposed algorithm re-organizes dataset into a form of nested binary tree*. Data items are compared at each node with only two nearest means with respect to each dimension and assigned to the one that has the closer mean. The main intuition of our research is as follows: We build the nested binary tree. Then we scan the data in raster order by in-order traversal of the tree. Lastly we compare data item at each node to the only two nearest means to assign the value to the intendant cluster. In this way we are able to save the computational cost significantly by reducing the number of comparisons with means and also by the least use to Euclidian distance formula. Our results showed that our method can perform clustering operation much faster than the classical ones. © Springer-Verlag Berlin Heidelberg 2005