GeneZilla

GeneZilla predicts complete protein-coding genes in human and other eukaryotic DNA sequences using graph-theoretic ORF representations, interpolated Markov models (IMMs), maximal dependence decomposition (MDD), and explicit biological feature states.

Key Features:

• Graph-Theoretic Representations: Employs graph-theoretic representations of high-scoring open reading frames (ORFs) to explore sub-optimal gene models beyond the single best prediction.
• Interpolated Markov Models (IMMs) and Maximal Dependence Decomposition (MDD): Uses IMMs and MDD to model sequence composition and dependencies for improved coding-region prediction.
• Biological Feature States: Models specific states for signal peptides, branch points, TATA boxes, and CAP sites, with planned incorporation of CpG islands.
• Evidence Combiner: Integrates additional evidence tracks via an evidence-based "combiner" program, which has shown substantial accuracy improvements when supplied with more evidence.
• Implementation and Adaptability: Implemented in C++ with flexible modeling strategies to accommodate different eukaryotic genome classes.

Scientific Applications:

• Controlled gene finding experiments: Applied in controlled experiments such as analyses of ENCODE regions to evaluate gene prediction approaches.
• Comparative evaluation of modeling strategies: Facilitates comparison of different strategies for modeling and predicting eukaryotic gene structures.
• Assessment of evidence integration: Used to assess the impact of incorporating diverse evidence tracks and feature states on prediction accuracy.
• Adaptation to novel genomes: Suited to analyses that require methodological adjustments for new classes of genomes.

Methodology:

Uses graph-theoretic ORF representations, interpolated Markov models (IMMs), maximal dependence decomposition (MDD), explicit states for signal peptides/branch points/TATA boxes/CAP sites (with planned CpG islands), an evidence-based "combiner" program, and pipeline adaptations and controlled experiments to accommodate human DNA and assess the impact of added evidence and feature states.