IRCAS: a novel end-to-end approach to identify, rectify, and classify comprehensive alternative splicing events in a...

Area of Science:

• Genomics
• Bioinformatics
• Molecular Biology

Background:

• Alternative splicing (AS) is crucial for proteomic diversity and adaptation in eukaryotes.
• Current AS detection methods require reference genomes, limiting their use in non-model organisms.
• Existing reference-free methods have inaccurate splice site prediction and separate detection/classification steps.

Purpose of the Study:

• To introduce IRCAS, an integrated end-to-end framework for reference-free alternative splicing analysis.
• To overcome limitations of existing methods in non-model organisms.
• To improve accuracy and efficiency in AS detection and classification.

Main Methods:

• IRCAS utilizes colored de Bruijn graphs for AS detection.
• An attention-based convolutional neural network is employed for splice site rectification.
• A hybrid graph neural network (graph attention network and Transformer layers) is used for classification.

Main Results:

• Splice site accuracy improved to 92%-96% compared to 50%-55% for existing methods.
• End-to-end inference accuracy reached 83.4% in rice, significantly outperforming the previous best method (44.7%).
• IRCAS demonstrated substantial improvements across four tested species.

Conclusions:

• IRCAS establishes a new benchmark for reference-free AS detection.
• The framework enhances the analysis of alternative splicing in non-model organisms.
• IRCAS offers a more accurate and integrated approach to AS analysis.