New feature subset selection procedures for classification of expression profiles

Methods for extracting useful information from the datasets produced by microarray experiments are at present of much interest. Here we present new methods for finding gene sets that are well suited for distinguishing experiment classes, such as healthy versus diseased tissues. Our methods are based on evaluating genes in pairs and evaluating how well a pair in combination distinguishes two experiment classes. We tested the ability of our pair-based methods to select gene sets that generalize the differences between experiment classes and compared the performance relative to two standard methods. To assess the ability to generalize class differences, we studied how well the gene sets we select are suited for learning a classifier. We show that the gene sets selected by our methods outperform the standard methods, in some cases by a large margin, in terms of cross-validation prediction accuracy of the learned classifier. We show that on two public datasets, accurate diagnoses can be made using only 15-30 genes. Our results have implications for how to select marker genes and how many gene measurements are needed for diagnostic purposes. When looking for differential expression between experiment classes, it may not be sufficient to look at each gene in a separate universe. Evaluating combinations of genes reveals interesting information that will not be discovered otherwise. Our results show that class prediction can be improved by taking advantage of this extra information.     

Structure motif discovery and mining the PDB

MOTIVATION: Many of the most interesting functional and evolutionary relationships among proteins are so ancient that they cannot be reliably detected through sequence analysis and are apparent only through a comparison of the tertiary structures. The conserved features can often be described as structural motifs consisting of a few single residues or Secondary Structure (SS) elements. Confidence in such motifs is greatly boosted when they are found in more than a pair of proteins. RESULTS: We describe an algorithm for the automatic discovery of recurring patterns in protein structures. The patterns consist of individual residues having a defined order along the protein's backbone that come close together in the structure and whose spatial conformations are similar. The residues in a pattern need not be close in the protein's sequence. The work described in this paper builds on an earlier reported algorithm for motif discovery. This paper describes a significant improvement of the algorithm which makes it very efficient. The improved efficiency allows us to use it for doing unsupervised learning of patterns occurring in small subsets in a large set of structures, a non-redundant subset of the Protein Data Bank (PDB) database of all known protein structures.     

Biodistribution of long-circulating PEG-liposomes in a murine model of established subcutaneous abscesses

The biodistribution of long-circulating PEG-liposomes in a subcutaneous mouse model of established mixed infection abscesses was investigated to assess their possible role as drug carriers in the treatment of small, undrainable intra-abdominal abscesses. There was a 10-30-fold greater localisation of (67)Ga-labelled PEG-liposomes in abscesses compared to uninfected normal skin samples. Over 3% of the injected dose (ID) of liposomes was present in the abscesses 24 h after liposome administration in contrast to 0.1% in normal skin sections. The percentage ID present in the liver, spleen and kidneys was 17%, 4% and 2% per organ respectively. Five days after liposome injection, 2% ID could still be recovered from the abscesses. Using colloidal gold-labelled PEG-liposomes, it was shown that there was a 4-fold greater density of liposome clusters in the subcutaneous tissue surrounding the capsule than in the core of the abscesses. The clusters within the abscesses were distributed evenly. We conclude that PEG-liposomes localise to a significant degree at the infection focus in our mouse model and may provide a new approach to the antimicrobial treatment of intra-abdominal abscesses.     

Development of Cuscuta species on a partially incompatible host: induction of xylem transfer cells

Summary.  The growth of dodders, Cuscuta reflexa and Cuscuta japonica, on the partially incompatible host poinsettia (Euphorbia pulcherrima) is studied. Poinsettia responds by bark growths to the formation of the dodder haustoria and prevents dodder from obtaining normal growth. The growth instead becomes extremely branched, coral-like, and dodder lacks the ability to form haustoria. After a period of coral-like growth, long shoots sprout, resembling the normal growth. These long shoots mark an ending phase for dodder, which dies shortly after without having flowered. During the coral-like growth phase, dodder develops transfer cells in the parenchyma cells bordering the vessels of the xylem in the shoot. The transfer cells have not been observed when dodder is grown on the compatible host Pelargonium zonale. A coral-like growth phase has also been observed at the establishing phase when dodder is grown in vitro on agar; later a more normal growth form takes over. In this coral phase, xylem transfer cells are also developed. The fluorochromes carboxyfluorescein and Texas Red were loaded into the host in the phloem and xylem, respectively, and detection of these fluorochromes in the dodder stem indicated that a functional haustorial contact developed for both vascular systems. The results show that Cuscuta spp. have the genetic ability to develop xylem transfer cells and use this in response to developmental stress. Received June 12, 2002; accepted August 26, 2002; published online March 11, 2003     

Visualization of the target-membrane-inserted fusion protein of Semliki Forest virus by combined electron microscopy and crystallography

Semliki Forest virus enters cells by receptor-mediated endocytosis. The acidic environment of the endosome triggers a membrane fusion reaction that is mediated by the E1 glycoprotein. During fusion, E1 rearranges from an E1/E2 heterodimer to a highly stable, membrane-inserted E1 homotrimer (E1HT). In this study, we analyzed E1HT by a combination of electron cryomicroscopy, electron crystallography of negatively stained 2D crystals, and fitting of the available X-ray structure of the monomeric E1 ectodomain into the resulting 3D reconstruction. The visualized E1HT reveals that the ectodomain has reoriented vertically and inserted the distal tip of domain II into the lipid bilayer. Our data allow the visualization of a viral fusion protein inserted in its target membrane and demonstrate that insertion is a cooperative process, resulting in rings composed of five to six homotrimers.     

Effect of bradykinin analogues on the B1 receptor of rat ileum

The longitudinal muscle of isolated rat ileum is a sensitive bioassay suitable for testing compounds with antagonistic effects on the B(1) receptor. Bradykinin analogues with replacement of proline by alkyl-substituted phenylalanine at position 7 are effective on this receptor as entire molecules and have a stronger antagonistic effect than on the B(2) receptor. A corresponding desArg(9)-compound has a specific effect on the B(1) receptor and a very high antagonistic potency. [LNMPhe(2)]bradykinin as a compound without any replacement at position 7 or 8 shows antagonistic activity as well.     

Sequence, distribution and quantification of the motilin precursor in the cat

Due to motilin's relation to the migrating motor complex (MMC), the physiology of motilin has been mostly studied in man and dog. The cat does not have an MMC pattern, and little is known about cat motilin. Therefore we identified the cat motilin precursor (GenBank accession no. AF127917) and developed a quantitative polymerase chain reaction (PCR) to explore its distribution in the gastrointestinal tract and in the central nervous system (CNS). The precursor is closely related to the dog precursor and consists of an open reading frame of 348bp encoding the signal peptide (25 amino acids), the motilin sequence (22 amino acids) and the motilin associated peptide (69 amino acids). One amino acid of the signal peptide was subject to gene polymorphism. Quantification of motilin messenger RNA (mRNA) was for the first time achieved. It is most abundant in the gastrointestinal tract, with the highest concentration in the duodenum, the lowest in the colon and is not detectable in the corpus. However an important expression was also observed in several regions of the CNS, except the striatum and cerebral cortex. The highest level was in the hypothalamus (although 23-fold lower than in the duodenum), the lowest level in the pons. Moderate levels were found in the thyroid. These data suggest that the physiological role of motilin may extend beyond its effect on gastrointestinal motility.     

Apparently normal tumor necrosis factor receptor 1 signaling in the absence of the silencer of death domains

The silencer of death domains (SODD) has been proposed to prevent constitutive signaling of tumor necrosis factor receptor 1 (TNFR1) in the absence of ligand. Besides TNFR1, death receptor 3 (DR3), Hsp70/Hsc70, and Bcl-2 have been characterized as binding partners of SODD. In order to investigate the in vivo role of SODD, we generated mice congenitally deficient in expression of the sodd gene. No spontaneous inflammatory infiltrations were observed in any organ of these mice. Consistent with this finding, in the absence of SODD no alteration in the activation patterns of nuclear factor kappaB (NF-kappaB), stress kinases, or ERK1 or -2 was observed after stimulation with tumor necrosis factor (TNF). Activation of NF-kappaB by DR3 was also unchanged. The extents of DR3- and TNF-induced apoptosis were comparable in gene-deficient and wild-type cells. Protection of cells against heat shock as mediated by the Hsp70 system and against staurosporine-induced apoptosis was independent of SODD. Furthermore, resistance to high-dose lipopolysaccharide (LPS) injections, LPS-D-GalN injections, and infection with listeriae was similar in wild-type and gene-deficient mice. In conclusion, our data do not support the concept of a unique, nonredundant role of SODD for the functions of TNFR1, Hsp70, and DR3.     

The Gap1 general amino acid permease acts as an amino acid sensor for activation of protein kinase A targets in the yeast Saccharomyces cerevisiae

Addition of a nitrogen source to yeast (Saccharomyces cerevisiae) cells starved for nitrogen on a glucose-containing medium triggers activation of protein kinase A (PKA) targets through a pathway that requires for sustained activation both a fermentable carbon source and a complete growth medium (fermentable growth medium induced or FGM pathway). Trehalase is activated, trehalose and glycogen content as well as heat resistance drop rapidly, STRE-controlled genes are repressed, and ribosomal protein genes are induced. We show that the rapid effect of amino acids on these targets specifically requires the general amino acid permease Gap1. In the gap1Delta strain, transport of high concentrations of l-citrulline occurs at a high rate but without activation of trehalase. Metabolism of the amino acids is not required. Point mutants in Gap1 with reduced or deficient transport also showed reduced or deficient signalling. However, two mutations, S391A and S397A, were identified with a differential effect on transport and signalling for l-glutamate and l-citrulline. Specific truncations of the C-terminus of Gap1 (e.g. last 14 or 26 amino acids) did not reduce transport activity but caused the same phenotype as in strains with constitutively high PKA activity also during growth with ammonium as sole nitrogen source. The overactive PKA phenotype was abolished by mutations in the Tpk1 or Tpk2 catalytic subunits. We conclude that Gap1 acts as an amino acid sensor for rapid activation of the FGM signalling pathway which controls the PKA targets, that transport through Gap1 is connected to signalling and that specific truncations of the C-terminus result in permanently activating Gap1 alleles.     

Humic substances in drinking water and the epidemiology of thyroid disease

Thyroid diseases are common in all populations but the type and frequency depends on environmental factors. In Denmark geographical differences in iodine intake are caused by different iodine contents of drinking water, which varies from < 1 to 139 microg iodine per litre. Comparative epidemiologic studies have demonstrated considerable differences in type and occurrence of thyroid disease with more goitre and hyperthyroidism in Aalborg with water iodine content around 5 microg/L, and more hypothyroidism in Copenhagen with water iodine around 20 microg/L. In Denmark, iodine in ground water is bound in humic substances, which have probably leached from marine sediments in the aquifers. Interestingly, humic substances in water from other parts of the world have goitrogenic properties, especially humic substances from coal and shale. Humic substances are heterogeneous mixtures of naturally occurring molecules, produced by decomposition of plant and animal tissues. The effect of humic substances in drinking water on the epidemiology of thyroid disease probably depends on the source of aquifer sediments.