SPA: a probabilistic algorithm for spliced alignment.

Area of Science:

• Computational Biology
• Genomics
• Bioinformatics

Background:

• Alternative splicing generates significant proteome diversity in higher eukaryotes.
• Accurate cDNA-to-genome mapping is essential for studying splice variants and mechanisms.
• Existing alignment algorithms lack the accuracy needed due to suboptimal scoring models.

Purpose of the Study:

• To develop a more accurate computational method for cDNA-to-genome alignment.
• To improve the identification of splice variants and understand alternative splicing mechanisms.
• To enable more precise mapping of transcript starts, ends, and subtle splice variations.

Main Methods:

• Developed a Bayesian probabilistic approach for cDNA-to-genome alignment.
• Incorporated prior probabilities for gene structures based on intron/exon lengths and splice boundaries.
• Utilized a likelihood model accounting for various sequencing error types and rates.

Main Results:

• Implemented the method in the SPA (Splicing Prediction Algorithm) program.
• SPA demonstrated significantly higher alignment quality compared to four other programs on large human and mouse cDNA datasets.
• SPA achieved superior accuracy in mapping splice boundaries and transcript 5'/3' ends, identifying subtle splice variations.

Conclusions:

• The Bayesian approach and SPA software provide a substantial improvement in cDNA-to-genome alignment accuracy.
• Enhanced mapping facilitates more accurate studies of alternative splicing and proteome diversity.
• Analysis revealed a potential novel non-canonical splice site conserved across human and mouse.