NetPlantGene

NetPlantGene predicts intron splice sites in the DNA of the dicot plant Arabidopsis thaliana by combining artificial neural networks with a rule-based system to improve splice site annotation and reduce false positives.

Key Features:

• Artificial neural network integration: Combines artificial neural networks with a rule-based system to generate splice site predictions and reduce false positives.
• Two-step prediction scheme: Performs an initial global assessment of coding potential that informs a subsequent local prediction of splice sites.
• Refined cutoff determination: Determines refined cutoffs based on splice site confidence values, prediction scores, and the coding context.
• Rule-based refinement: Applies rules that consider distances between potential splice sites and other contextual information to refine predictions.
• Non-local interaction modeling: Incorporates non-local interactions among predicted splice sites to enhance overall accuracy.
• Error analysis and novel discovery: Analysis of prediction errors revealed frequent T-tract prolongation containing cryptic acceptor sites at the 5' end of exons.
• Comparative superiority: Outperforms methods such as GeneFinder, Gene-Mark, and Grail by an order of magnitude.
• Experimental validation: Identified a donor splice site within the coding sequence for the jellyfish Green Fluorescent Protein at the exact position observed in A. thaliana transformants.
• Application to alternative splicing: Provides insights into alternatively spliced genes and alternative splicing events.
• Broad applicability: Demonstrates examples of applicability to other dicots, monocots, and algae.

Scientific Applications:

• Gene annotation: Facilitates annotation of gene structures by predicting donor and acceptor splice sites in plant genomes.
• Alternative splicing analysis: Enables investigation of alternative splicing mechanisms and isoform variation in Arabidopsis thaliana.
• Gene structure and function studies: Supports studies of gene structure and function in plants by providing splice site evidence.
• Comparative genomics: Allows exploration of splice site patterns across dicots, monocots, and algae.
• Transgene splicing validation: Assists validation of transgene splice sites, as exemplified by GFP donor site identification in A. thaliana transformants.
• Genetic diversity exploration: Aids exploration of genetic diversity related to splice site variation.

Methodology:

Combines artificial neural networks with a rule-based system using a two-step scheme—an initial global coding-potential assessment followed by local splice-site prediction—with refined cutoffs based on splice site confidence, prediction scores and coding context, and rule-based refinements that consider distances and non-local interactions among predicted splice sites; prediction-error analysis identified T-tract prolongation with cryptic acceptor sites.