Two recent science paper (PMID:23258891 & 23258890) access evolutionary changes of alternative splicing. For most of the time, they relied on PSI (percentage spliced in) of orthologous exons. But there's some obvious problems:

1. PSI is relative to the most frequent exon (or say "exon junction") in the gene. But "most frequent" does not mean constitutive (always used). Imagine a complex gene that no exon is constitutive, PSI is then underestimating what it suppose to measure.

2. PSI does not reflect the connection between exons. For example, an exon could have 100% PSI in both human and mouse, but in human, it always connect with another in-frame exon, but in mouse it always connect with another out-of-frame exon, making the 100% PSI functionless in mouse.

May be there's other problems in PSI, but as far as I can think of, these two are the ones that could significantly biased the conclusions they made in these two Science papers.

So my question is, is there other computational methods to quantitatively compare alternative splicing? If you have some suggestion on experimental improvements, you're also welcomed to post it here.

Thank you all !

Article The Evolutionary Landscape of Alternative Splicing in Verteb...