Spliceator

Spliceator predicts eukaryotic splice sites using convolutional neural networks to support gene annotation across diverse species.

Key Features:

• Deep learning (CNN): Uses convolutional neural networks to predict both donor and acceptor splice sites from genomic sequences.
• Broad species applicability: Trained on validated data from over 100 different organisms, enabling predictions for non-model and model species.
• High predictive accuracy: Achieves 89–92% performance on independent benchmarks across taxa including humans, fish, flies, worms, plants, and protists.
• Data-driven training: Model training relies on high-quality, validated cross-species splice site datasets to improve generalization.

Scientific Applications:

• Genome annotation: Predicts splice sites to support eukaryotic gene annotation and exon-intron boundary identification.
• Gene structure identification: Assists in delineating donor and acceptor sites to define gene structures in both model and non-model organisms.
• Comparative genomics: Enables splice site prediction across diverse taxa to facilitate comparative studies of genomic organization and function.

Methodology:

Spliceator trains convolutional neural networks on validated cross-species splice site datasets derived from over 100 organisms.