Classification and Categorical Inputs with Treed Gaussian Process Models

Recognizing the successes of treed Gaussian process (TGP) models as an interpretable and thrifty model for nonparametric regression, we seek to extend the model to classification. Both treed models and Gaussian processes (GPs) have, separately, enjoyed great success in application to classification problems. An example of the former is Bayesian CART. In the latter, real-valued GP output may be utilized for classification via latent variables, which provide classification rules by means of a softmax function. We formulate a Bayesian model averaging scheme to combine these two models and describe a Monte Carlo method for sampling from the full posterior distribution with joint proposals for the tree topology and the GP parameters corresponding to latent variables at the leaves. We concentrate on efficient sampling of the latent variables, which is important to obtain good mixing in the expanded parameter space. The tree structure is particularly helpful for this task and also for developing an efficient scheme for handling categorical predictors, which commonly arise in classification problems. Our proposed classification TGP (CTGP) methodology is illustrated on a collection of synthetic and real data sets. We assess performance relative to existing methods and thereby show how CTGP is highly flexible, offers tractable inference, produces rules that are easy to interpret, and performs well out of sample.     

Common variation at 10p12.31 near MLLT10 influences meningioma risk

To identify susceptibility loci for meningioma, we conducted a genome-wide association study of 859 affected individuals (cases) and 704 controls with validation in two independent sample sets totaling 774 cases and 1,764 controls. We identified a new susceptibility locus for meningioma at 10p12.31 (MLLT10, rs11012732, odds ratio = 1.46, P(combined) = 1.88 × 10(-14)). This finding advances our understanding of the genetic basis of meningioma development.     

Sequence of Plasmodium falciparum chromosomes 2, 10, 11 and 14

The mosquito-borne malaria parasite Plasmodium falciparum kills an estimated 0.7-2.7 million people every year, primarily children in sub-Saharan Africa. Without effective interventions, a variety of factors-including the spread of parasites resistant to antimalarial drugs and the increasing insecticide resistance of mosquitoes-may cause the number of malaria cases to double over the next two decades. To stimulate basic research and facilitate the development of new drugs and vaccines, the genome of Plasmodium falciparum clone 3D7 has been sequenced using a chromosome-by-chromosome shotgun strategy. We report here the nucleotide sequences of chromosomes 10, 11 and 14, and a re-analysis of the chromosome 2 sequence. These chromosomes represent about 35% of the 23-megabase P. falciparum genome.     

Low-penetrance susceptibility to breast cancer due to CHEK2(*)1100delC in noncarriers of BRCA1 or BRCA2 mutations

Mutations in BRCA1 and BRCA2 confer a high risk of breast and ovarian cancer, but account for only a small fraction of breast cancer susceptibility. To find additional genes conferring susceptibility to breast cancer, we analyzed CHEK2 (also known as CHK2), which encodes a cell-cycle checkpoint kinase that is implicated in DNA repair processes involving BRCA1 and p53 (refs 3,4,5). We show that CHEK2(*)1100delC, a truncating variant that abrogates the kinase activity, has a frequency of 1.1% in healthy individuals. However, this variant is present in 5.1% of individuals with breast cancer from 718 families that do not carry mutations in BRCA1 or BRCA2 (P = 0.00000003), including 13.5% of individuals from families with male breast cancer (P = 0.00015). We estimate that the CHEK2(*)1100delC variant results in an approximately twofold increase of breast cancer risk in women and a tenfold increase of risk in men. By contrast, the variant confers no increased cancer risk in carriers of BRCA1 or BRCA2 mutations. This suggests that the biological mechanisms underlying the elevated risk of breast cancer in CHEK2 mutation carriers are already subverted in carriers of BRCA1 or BRCA2 mutations, which is consistent with participation of the encoded proteins in the same pathway.     

The reversibility of mitotic exit in vertebrate cells

A guiding hypothesis for cell-cycle regulation asserts that regulated proteolysis constrains the directionality of certain cell-cycle transitions. Here we test this hypothesis for mitotic exit, which is regulated by degradation of the cyclin-dependent kinase 1 (Cdk1) activator, cyclin B. Application of chemical Cdk1 inhibitors to cells in mitosis induces cytokinesis and other normal aspects of mitotic exit, including cyclin B degradation. However, chromatid segregation fails, resulting in entrapment of chromatin in the midbody. If cyclin B degradation is blocked with a proteasome inhibitor or by expression of non-degradable cyclin B, Cdk inhibitors will nonetheless induce mitotic exit and cytokinesis. However, if after mitotic exit, the Cdk1 inhibitor is washed free from cells in which cyclin B degradation is blocked, the cells can revert back to M phase. This reversal is characterized by chromosome recondensation, nuclear envelope breakdown, assembly of microtubules into a mitotic spindle, and in most cases, dissolution of the midbody, reopening of the cleavage furrow, and realignment of chromosomes at the metaphase plate. These findings demonstrate that proteasome-dependent degradation of cyclin B provides directionality for the M phase to G1 transition.     

No supernovae associated with two long-duration gamma-ray bursts

It is now accepted that long-duration gamma-ray bursts (GRBs) are produced during the collapse of a massive star. The standard 'collapsar' model predicts that a broad-lined and luminous type Ic core-collapse supernova accompanies every long-duration GRB. This association has been confirmed in observations of several nearby GRBs. Here we report that GRB 060505 (ref. 10) and GRB 060614 (ref. 11) were not accompanied by supernova emission down to limits hundreds of times fainter than the archetypal supernova SN 1998bw that accompanied GRB 980425, and fainter than any type Ic supernova ever observed. Multi-band observations of the early afterglows, as well as spectroscopy of the host galaxies, exclude the possibility of significant dust obscuration and show that the bursts originated in actively star-forming regions. The absence of a supernova to such deep limits is qualitatively different from all previous nearby long-duration GRBs and suggests a new phenomenological type of massive stellar death.     

Specific structural features of LnZrOx (Ln: La, Sm) mixed oxides prepared by different methods

LnZrOx(Ln: La, Sm) mixed oxides of Ln: Zr = 1 were prepared by different methods(complex polymerized method, sorption of cations on starch from aqueous salt solution and conventional co-precipitation with additional redispersion of precipitate by ultra sound) and calcined at 700–1300 °C. Their specific structural features and changes were studied and discussed. Various characterization methods were used such as X-ray diffraction,Electron microscopy, Fourier-transform infrared and Raman spectroscopy, UV–Vis spectroscopy, X-Ray absorption fine structure and X-ray photoelectron spectroscopy.The formation of pyrochlore structure occurred at 1100–1300 °C from fluorite-like pseudocubic phase ZrO_2 regardless the method of preparation. This phase had a block-like structure consisting of ZrO_2 nanocrystals stabilized by Ln cations and residual anions such as hydroxyls and carbonates. The desorption of such anions with heating already started at 900 °C and lead to local changes of Zr cations coordination to octahedral and to the formation of pyrochlore nanodomains inclusions within fluorite-like phase. The increased cation mobility and further elimination of anions caused by further heating was accompanied by the formation of bulk pyrochlore phase at 1100–1300 °C. Even after calcination at 1300 °C the local microheterogeneity as well as defects were identified at domains boundaries or sintered microstructure. These specific features of the formed pyrochlores depend on the method of preparation.     

Evidence for two independent prostate cancer risk-associated loci in the HNF1B gene at 17q12

We carried out a fine-mapping study in the HNF1B gene at 17q12 in two study populations and identified a second locus associated with prostate cancer risk, approximately 26 kb centromeric to the first known locus (rs4430796); these loci are separated by a recombination hot spot. We confirmed the association with a SNP in the second locus (rs11649743) in five additional populations, with P = 1.7 x 10(-9) for an allelic test of the seven studies combined. The association at each SNP remained significant after adjustment for the other SNP.