Search

A method for de novo genome annotation using high-throughput cDNA sequencing data., Next generation technologies enable massive-scale cDNA sequencing (so-called RNA-Seq). Mainly because of the difficulty of aligning short reads on exon-exon junctions, no attempts have been made so far to use RNA-Seq for building gene models de novo, that is, in the absence of a set of known genes and/or splicing events. We present G-Mo.R-Se (Gene Modelling using RNA-Seq), an approach aimed at building gene models directly from RNA-Seq and demonstrate its utility on the grapevine genome.

The genome of Saccharomyces kluyveri was explored through 2528 random sequence tags with an average length of 981 bp. The complete nuclear ribosomal DNA unit was found to be 8656 bp in length. Sequences homologous to retroelements of the gypsy and copia types were identified as well as numerous solo long terminal repeats. We identified at least 1406 genes homologous to Saccharomyces cerevisiae open reading frames, with on average 58.1% and 72.4% amino acid identity and similarity, respectively. In addition, by comparison with completely sequenced genomes and the SwissProt database, we found 27 novel S. kluyveri genes. Most of these genes belong to pathways or have functions absent from S. cerevisiae , such as the catabolic pathway of purines or pyrimidines, melibiose fermentation, sorbitol utilization, or degradation of pollutants. The sequences are deposited in EMBL under the accession numbers AL404849 – AL407376 .

As part of a comparative genomics project on 13 hemiascomycetous yeasts, the Pichia angusta type strain was studied using a partial random sequencing strategy. With coverage of 0.5 genome equivalents, about 2500 novel protein-coding genes were identified, probably corresponding to more than half of the P. angusta protein-coding genes, 6% of which do not have homologs in Saccharomyces cerevisiae. Some of them contain one or two introns, on average three times shorter than those in S. cerevisiae. We also identified 28 tRNA genes, a few retrotransposons similar to Ty5 of S. cerevisiae, solo long terminal repeats, the whole ribosomal DNA cluster, and segments of mitochondrial DNA. The P. angusta sequences were deposited in EMBL under the accession numbers AL430961 to AL436044.

Comparisons of the 6213 predicted Saccharomyces cerevisiae open reading frame (ORF) products with sequences from organisms of other biological phyla differentiate genes commonly conserved in evolution from ‘maverick’ genes which have no homologue in phyla other than the Ascomycetes. We show that a majority of the ‘maverick’ genes have homologues among other yeast species and thus define a set of 1892 genes that, from sequence comparisons, appear ‘Ascomycetes-specific’. We estimate, retrospectively, that the S. cerevisiae genome contains 5651 actual protein-coding genes, 50 of which were identified for the first time in this work, and that the present public databases contain 612 predicted ORFs that are not real genes. Interestingly, the sequences of the ‘Ascomycetes-specific’ genes tend to diverge more rapidly in evolution than that of other genes. Half of the ‘Ascomycetes-specific’ genes are functionally characterized in S. cerevisiae, and a few functional categories are over-represented in them.

We have evaluated the degree of gene redundancy in the nuclear genomes of 13 hemiascomycetous yeast species. Saccharomyces cerevisiae singletons and gene families appear generally conserved in these species as singletons and families of similar size, respectively. Variations of the number of homologues with respect to that expected affect from 7 to less than 24% of each genome. Since S. cerevisiae homologues represent the majority of the genes identified in the genomes studied, the overall degree of gene redundancy seems conserved across all species. This is best explained by a dynamic equilibrium resulting from numerous events of gene duplication and deletion rather than by a massive duplication event occurring in some lineages and not in others.

The genome of the diploid hemiascomycetous yeast Candida tropicalis, an opportunistic human pathogen and an important organism for industrial applications, was explored by the analysis of 2541 Random Sequenced Tags (RSTs) covering about 20% of its genome. Comparison of these sequences with Saccharomyces cerevisiae and other species permitted the identification and the analysis of a total of more than 1000 novel genetic elements of C. tropicalis. Moreover, the present study confirms that in C. tropicalis, the rare CUG codon is read as a serine and not a leucine. The sequences have been deposited at EMBL with the accession numbers AL438875–AL441602.

This paper reports the genomic analysis of the strain CBS7064 of Pichia sorbitophila, a homothallic diploid yeast. We sequenced 4829 random sequence tags of about 1 kb and compared them to the Saccharomyces cerevisiae gene products. Approximately 1300 nuclear genes, 22 tRNAs, the rDNA locus, and six mitochondrial genes have been identified. The analysis of the rDNA genes has permitted to classify this organism close to the Candida species. Accession numbers from AL414896 to AL419724 at EMBL databank.

Interfaces de interações proteína-proteína contêm informações importantes sobre o reconhecimento molecular. Este trabalho versa sobre a descoberta de padrões conservados nesta região. Nós entendemos que é essencial compreender como substratos e inibidores são ligados ou para predizer a ligação molecular. Quando um inibidor liga-se a enzimas com estruturas, sequências ou mecanismos diferentes, nós chamamos esta falta de especificidade de Inibição Cruzada. A identicação de variantes é um tarefa difícil no qual os métodos tradicionais podem falhar. Para entender como a inibição cruzada ocorre, nós modelamos o problema e propomos uma metodologia, na qual chamamos de HydroPaCe (Hydrofobic Patch Centroid ) e avaliamos, por meio de métrica de cobertura penalizada PRM (Penalized Recall Metric), para detectar padrões conservados. Os átomos hidrof óbicos que compõem a interface foram modelados como grafos de interações apolares e os patches ou regiões conectadas foram computadas e sumarizadas em centroides geométricos (HP-centroides), e sua conservação foi detectada. Nós analisamos dois casos de inibição cruzada, sendo um com o inibidor Ovomovoide (5 complexos não redundantes) e outro com o inibidor Eglina C (4 complexos não redundantes). Além disso, duas famílias de serino proteases, as Tipo Tripsina (35 estruturas não redundantes) e Tipo Subtilisina (9 estruturas não redundantes), com cadeias simples foram analisadas, usando técnica de projeção da interface. Os padrões encontrados nas interfaces das inibições cruzadas foram encontrados na maioria das enzimas destas famílias. Acreditamos que nossa metodologia atingiu um nível de abstração para obter propriedades invariantes em inibição cruzada. Nós mostramos exemplos onde os HP-centroides foram preditos com sucesso em enzimas por inibidores já estudados, de acordo com a base de dados BRENDA. Por fim, nós disponibilizamos todos os códigos, usados nos passos metodológicos, bem como, tutoriais e exemplos em www.dcc.ufmg.br/raquelcm/hydropace. Interfaces of protein-protein interactions contain important information about molecular recognition. This work aims at the discovery of conserved patterns in this region. We believe it is essential to understand how substrates and inhibitors are bound and for predicting molecular binding. When an inhibitor binds to enzymes with different structures, sequences or mechanisms, we call this lack of specicity of Cross-Inhibition. The identication of variants is a dicult task for which traditional methods may fail. To understand how the cross-inhibition occurs, we model the problem and propose a methodology, whichwe call HydroPaCe(HydrofobicPatch Centroid) and we evaluate, through penalized recall metric - PRM, to detect preserved patterns. The hydrophobic atoms that belong to the interface were modeled as graphs of apolar interactions and hydrofobic connected regions or patches were computed and summarized by geometric centroids (HP-centroids), andtheir conservation is detected. We analyze two cases of cross-inhibition, one with the inhibitor Ovomovoid (5 non-redundant complexes) and another with the inhibitor Eglin C (4 non-redundant complexes). In addition, two families of serine proteases with single chain (apo-enzymes) were analyzed using projection of the interface technique: The Trypsin-like (35 non-redundant structures) and the Subtilisin-like (9 non-redundant structures). The patterns found in the interfaces of cross-inhibitions were found in most enzymes these families. We believe that our methodology achieves an appropriate level of abstraction toobtain invariant properties in cross-inhibition. We show examples in which HP-centroids successfully predicted enzymes that could be inhibited by the studied inhibitors according to BRENDA database. Finally, we provide all codes used in the methodological steps, aswell as tutorials and examples in www.dcc.ufmg.br/raquelcm/hydropace.