Twinscan

Twinscan predicts gene structures in eukaryotic genomes by using comparative genomics to improve the accuracy of gene annotation.

Key Features:

• Comparative genomic approach: Predicts genes by comparing a target genome to a closely related genome and has been applied to mammals (human, mouse), Arabidopsis thaliana, Caenorhabditis elegans, C. briggsae, and Cryptococcus neoformans strains JEC21 and H99.
• Integration with GENSCAN: Extends the GENSCAN probability model to exploit homology between two related genomes.
• Separate models for genomic elements: Uses distinct probability models for exons, introns, splice sites, and untranslated regions (UTRs) to reflect differing conservation patterns.
• Conservation-based prediction: Incorporates conservation measurements across species to enhance identification of gene structures.
• High-throughput capability: Designed to process large genomic datasets with unknown numbers of genes for whole-genome analyses.
• Improved accuracy over previous methods: Demonstrates higher exon sensitivity and specificity and higher exact gene sensitivity and specificity compared to GENSCAN.
• Computational efficiency with sequencing intermediates: Leverages intermediate products—low-redundancy whole-genome shotgun reads, draft assemblies, and synteny maps—to maintain accuracy and efficiency, including performance on low coverage such as 1X mouse genome data.
• Conservative annotation strategy: Produces conservative gene sets in some applications, for example predicting 25,622 genes in human genome annotation using the mouse assembly.

Scientific Applications:

• Genome annotation: Facilitates comprehensive annotation of whole eukaryotic genomes using comparative evidence.
• Evolutionary studies: Enables analysis of evolutionary conservation and divergence of genomic elements across species.
• Functional genomics: Assists identification of functional genomic elements such as exons, introns, splice sites, and UTRs through comparative signals.

Methodology:

Extends GENSCAN's probability model with separate models for exons, introns, splice sites, and UTRs; performs pairwise comparative genomics between related genomes using conservation measurements; and uses intermediate sequencing products (low-redundancy whole-genome shotgun reads, draft assemblies, synteny maps) to support predictions, including low-coverage data (e.g., 1X mouse).