The Genom of Homo sapiens.pdf

WHOLE-GENOME COMPARISONS 277Figure 2. Construction of ecotigs. The two first lines represent, respectively, the ecores (boxes) and the HSPs (segments) detected ongenome B, using genome A as a query. The two following lines represent, respectively, the HSPs and the ecores detected on genomeA, using genome B as a query. The bottom line represents the ecotig gene models constructed on genome A. Matching HSPs are linkedby dotted lines. Matching ecores are identified by the same prefix (i, ii, etc.). Numbers over (or under) arrows represent distances separatingecores that are consecutive on genome A (number of consecutive ecores minus one).genes may remain colinear on the pair of comparedgenomes, depending on the degree of conserved syntenyof the genome pair analyzed. In the rest of this paper, wedesignate the pair of matched genomes with the followingnotation: (query/target).RESULTSThe ecore detection and ecotig construction procedureshave been applied to compare draft or complete genomesequences from various multicellular organisms. We appliedExofish comparisons to several genome pairs betweenplant, insect, and vertebrate genomes, namely,mammals (mouse/human) with pufferfish (Tetraodon/Takifugu), the Drosophila melanogaster sequence withthe Anopheles gambiae draft, and the Arabidopsis sequencewith the rice genome draft. Such comparisonswere used (1) to detect gene models or exons that havenot yet been identified in one or both genomes, (2) to extendexisting gene models, and (3) to determine the degreeof completion of existing annotations.Plant GenomesThe recent availability of a draft sequence of the ricegenome with sufficient coverage (Goff et al. 2002) hasopened the possibility of comparing whole plant genomesfor the first time. In addition, the genome of Arabidopsishas been the focus of several extensive annotation projectsthat make this genome one of the best documentedto date. This situation enabled us to use Arabidopsis as areference on which Exofish performances can be evalu-ated. Such an evaluation benefited also from the availabilityof a source of new cDNA sequences that have notyet been used in the Arabidopsis genome analyses buthave served as a support for experimental validation ofcomparison-based predictions.Exofish was first calibrated using a set of 1589 Arabidopsisgenes that had been manually annotated. The optimalconditions we determined produce a specificityabove 99% and a sensitivity at the exon and gene level of64% and 93%, respectively. The global Exofish comparisonwas performed between the finished Arabidopsisgenome sequence The Arabidopsis Genome Initiative(2000) used as a target and the BAC-based sequence draftestablished by the International Rice Genome SequencingProgram used as a query. Ecores were mapped relativeto the Arabidopsis genome annotation.Statistics on ecores detected within and outside annotationsare shown in Table 1. A total of 74% of the annotatedgenes (MIPS annotation) included one ecore atleast, and 47% of the annotated exons are matched by oneor more ecores. Conversely, 91% of the ecores are localizedwithin the boundaries of annotated genes, and onlyabout 1% of these ecores do not match an annotated exon.In a subset of 60 nonmatching ecores, experimental evidencebased on new cDNAs showed that 59 cases correspondto novel exons. We thus estimate that about 98% ofthe ecores within gene annotations, but which do notmatch annotated exons, correspond to real exons thatwere missed during the annotation process. Taking intoaccount that only one exon out of two is detected as anecore (Table 1), an extrapolation of this analysis suggeststhat about 900 internal exons are still missing in the set of