P0427Structural Gene Annotation of Fungal Genomes and Metagenomes
Structural Gene Annotation of Fungal Genomes and Metagenomes
Accurate gene annotation presents a challenge well known to the fungal genomics community. Structural gene annotation of complete or draft fungal genomes is difficult both in ‘short’ genomes, e.g. yeast, where the number of introns is too small to parametrize intron models, and in ‘long’ genomes, where large number of introns per gene presents, in fact, less issue than sequence composition heterogeneity and large populations of transposable elements. Finally, algorithms of structural annotation of fungal metagenomes simply do not exist.
We present here the first, to the best of our knowledge, method for gene identification in fungal metagenomes that takes advantage on availability of hundreds of already sequenced fungal genomes (Ding et al., in preparation). Development of this method was facilitated by experience gained in previous work of the gene prediction tool for prokaryotic metagenomes, MetaGeneMark (Zhu et al, 2010). The new algorithm uses G+C content of input sequence for selecting a model from a set of fungi kingdom specific heuristic models derived from more than 200 fungal genomes. The heuristic models also include G+C dependent sub-models for acceptor, donor and branch point sites. A subset of parameters, such as exon length distribution, was modeled with respect to bins with different intron density: low, medium and high.
For ab initio gene prediction in draft fungal genomes we have developed a group of fungi specific self-training methods. The constantly updated self- training tool GeneMark-ES has been used in a number of fungal genome sequencing projects since 2007. It remains to be a launch pad for integrated tools GeneMark-ET and GeneMark-EP that allow using mapped RNA-Seq reads and mapped homologous proteins, respectively, in training and prediction.
We present here the first, to the best of our knowledge, method for gene identification in fungal metagenomes that takes advantage on availability of hundreds of already sequenced fungal genomes (Ding et al., in preparation). Development of this method was facilitated by experience gained in previous work of the gene prediction tool for prokaryotic metagenomes, MetaGeneMark (Zhu et al, 2010). The new algorithm uses G+C content of input sequence for selecting a model from a set of fungi kingdom specific heuristic models derived from more than 200 fungal genomes. The heuristic models also include G+C dependent sub-models for acceptor, donor and branch point sites. A subset of parameters, such as exon length distribution, was modeled with respect to bins with different intron density: low, medium and high.
For ab initio gene prediction in draft fungal genomes we have developed a group of fungi specific self-training methods. The constantly updated self- training tool GeneMark-ES has been used in a number of fungal genome sequencing projects since 2007. It remains to be a launch pad for integrated tools GeneMark-ET and GeneMark-EP that allow using mapped RNA-Seq reads and mapped homologous proteins, respectively, in training and prediction.