Bayesian reconstruction and differential testing of excised introns.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Intron excision characterization is vital for understanding cellular complexity and disease.
• Current transcript reconstruction methods struggle with incomplete annotations and heterogeneous expression.
• Existing local splicing methods lack transcript-level information.

Purpose of the Study:

• To address limitations in transcript reconstruction and differential splicing analysis.
• To introduce a novel method that considers sequences of exon-exon junctions (SEEJs).
• To develop a Bayesian model for computing SEEJs and characterizing their differential usage.

Main Methods:

• Formalized a new transcript reconstruction problem using co-occurring SEEJs.
• Developed a hierarchical Bayesian admixture model (BSEEJ) for computing SEEJs.
• Employed a generalized linear model for differential SEEJ usage analysis.

Main Results:

• BSEEJ demonstrated high F1 scores in transcript reconstruction tasks.
• Achieved improved accuracy and sensitivity in differential splicing compared to existing methods on simulated data.
• Validated BSEEJ on experimental data for transcript reconstruction, novelty, and functional analysis.

Conclusions:

• BSEEJ offers a robust approach for transcript reconstruction and differential splicing analysis.
• The method effectively utilizes SEEJs to capture transcript-level splicing information.
• BSEEJ provides valuable insights into functional genomics and disease pathogenesis.