Abstract
Purpose::
Over half of human genes undergo alternative splicing (AS) and many retinal disease genes undergo tissue-specific AS. The aim of this study was to determine whether splice site (SS) strength scores, as computed by four bioinformatics algorithms, could predict AS in retinal transcripts.
Methods::
SS score was calculated on 1400 exons in 60 retinal dystrophy genes using four different scoring methods: Consensus Sequence Weighted Matrices, Neural Network, Information Theory and Maximum Entropy. Evidence of AS in each exon was obtained using the EST databases: Aceview and AltSplice. Based on splice site scoring and EST evidence, exons were classified into low and high SS scoring with or without evidence of AS. RT-PCR on human retinal RNA was performed to validate computational predictions in selected transcripts.
Results::
A higher percentage of low scoring exons, from all four computational scoring methods, are alternatively spliced according to EST data. RT-PCR confirmed that exons with high splice site scores when correlated with EST do not undergo AS . RT-PCR corroborated alternative splicing events in exons with low SS scores and EST evidence of AS. Low splice site score alone and EST evidence alone are poor predictors of AS.
Conclusions::
High SS scoring exons were found to rarely undergo AS, while there was variable AS in exons with lower SS score. Exons with low scoring splice sites and no EST or RT-PCR evidence of AS suggest the utilisation of cis-acting elements to regulate splicing. Splice site scores in conjunction with EST data can be used to either exclude or suggest the possibility of AS and thus facilitate an understanding of the functional diversity of splice variants in health and disease states.
Keywords: gene/expression • retina • transcription