P0276GeneMark Family of Gene Prediction Tools for Prokaryotic and Eukaryotic Genomes
GeneMark Family of Gene Prediction Tools for Prokaryotic and Eukaryotic Genomes
Gene prediction has a fundamental role in genomics. Still, the precise annotation of protein-coding genes in a new genomic sequence is the target difficult to reach.
Prokaryotic genes can be predicted with much higher average accuracy than eukaryotic ones. Nevertheless, the error rate is not negligible and largely species-specific. Most errors are made in prediction of genes located in genomic regions with atypical G+C composition. We present here results of further development of GeneMarkS, a self-training tool used in many genome projects. The new tool, GeneMarkS-2 (Tang et al. in preparation), uses local G+C-specific heuristic models for scoring individual ORFs in the first step of analysis. Predicted atypical genes are retained and serve as ‘external’ evidence in subsequent self-training iterations. GeneMarkS-2 controls the quality of training process by effectively computing a measure of relative entropy between competing sequence models parameterized by self-training, the measure that guides the selection of optimal model orders.
We also report progress in developing tools for structural annotation of eukaryotic genomes. We are constantly updating the self-training ab initio gene prediction tool, GeneMark-ES. This tool was extended to fully automated GeneMark-ET (Lomsadze et al., 2014) that integrates into the training process mapped RNA-Seq reads as well as to GeneMark-EP (Lomsadze et al. in preparation) that integrates information on footprints of homologous proteins. The automated tool that combines all the options, GeneMark-EX (paper in preparation) selects the mode of running depending on the availability and volume of the types of external evidence (transcripts and proteins).
Prokaryotic genes can be predicted with much higher average accuracy than eukaryotic ones. Nevertheless, the error rate is not negligible and largely species-specific. Most errors are made in prediction of genes located in genomic regions with atypical G+C composition. We present here results of further development of GeneMarkS, a self-training tool used in many genome projects. The new tool, GeneMarkS-2 (Tang et al. in preparation), uses local G+C-specific heuristic models for scoring individual ORFs in the first step of analysis. Predicted atypical genes are retained and serve as ‘external’ evidence in subsequent self-training iterations. GeneMarkS-2 controls the quality of training process by effectively computing a measure of relative entropy between competing sequence models parameterized by self-training, the measure that guides the selection of optimal model orders.
We also report progress in developing tools for structural annotation of eukaryotic genomes. We are constantly updating the self-training ab initio gene prediction tool, GeneMark-ES. This tool was extended to fully automated GeneMark-ET (Lomsadze et al., 2014) that integrates into the training process mapped RNA-Seq reads as well as to GeneMark-EP (Lomsadze et al. in preparation) that integrates information on footprints of homologous proteins. The automated tool that combines all the options, GeneMark-EX (paper in preparation) selects the mode of running depending on the availability and volume of the types of external evidence (transcripts and proteins).