GENSCAN

GENSCAN predicts gene structures in genomic DNA by applying a general probabilistic model that integrates transcriptional, translational, and splicing signals, length distributions, and compositional features to locate exons, introns, and intergenic regions.

Key Features:

• Probabilistic modeling: Employs a general probabilistic framework that accounts for gene density and structural variation across different C+G compositional regions of the human genome.
• Splice signal models: Models donor and acceptor splice signals and captures positional dependencies between signal positions.
• Signal and distribution integration: Integrates transcriptional, translational, and splicing signals with length distributions and compositional features of exons, introns, and intergenic regions.
• Region- and group-specific parameter sets: Derives distinct sets of model parameters tailored to genomic regions with varying C+G content and to different vertebrate groups.
• Gene identification (GENSCAN): Implemented as the GENSCAN program to identify complete exon/intron structures within genomic DNA.
• Multiple-gene and strand predictions: Predicts multiple genes per sequence, including partial and complete gene models, on one or both DNA strands.
• Confidence scoring and accuracy: Provides per-exon confidence indicators and identifies approximately 75 to 80% of exons exactly.

Scientific Applications:

• Gene prediction: Predicts exon–intron structures and locates multiple genes within genomic sequences.
• Strand-specific predictions: Generates consistent sets of gene predictions on either or both DNA strands.
• Prediction confidence and benchmarking: Supports assessment of prediction reliability using per-exon confidence indicators and high exon-level accuracy (75 to 80% exact identification).

Methodology:

Derives distinct parameter sets for different C+G compositional regions and applies a probabilistic model that integrates transcriptional, translational, and splicing signals, length distributions, compositional features, and splice-site dependency models.