CRAIG

CRAIG predicts gene structures in genomic sequences using a conditional random field (CRF)–based ab initio gene prediction framework.

Key Features:

• Conditional Random Field Gene Modeling: Uses a semi-Markov conditional random field (CRF) model to represent gene structures and dependencies among genomic features.
• Large-Margin Training Algorithm: Applies an online large-margin learning algorithm related to multiclass support vector machines (SVMs) to train prediction models.
• Integration of Genomic Evidence: Incorporates multiple genomic signals and sequence features to improve prediction of gene structures.
• Improved Long Intron Prediction: Demonstrates enhanced performance in predicting genes containing long introns.

Scientific Applications:

• Genome Annotation: Identifies gene structures in genomic sequences for annotation of eukaryotic genomes.
• Computational Genomics: Supports analysis of genomic regions such as vertebrate genomes and ENCODE project datasets.
• Gene Structure Analysis: Enables investigation of exon–intron organization and complex gene architectures.

Methodology:

CRAIG models gene structures using a semi-Markov conditional random field trained with an online large-margin algorithm related to multiclass support vector machines to integrate genomic sequence features and predict gene annotations.