We downloaded all the annotated internal human exons from ASAPII database [20] and detected those exons involved in tandem
bendamustine by examining the sequence similarity between adjacent exons (see methods section for details). If the “duplication-then-alternative” model was a prevalent phenomenon during vertebrate evolution, we would see a higher percentage of AS in the duplicated exons than in non-duplicated ones. The percentages of AS in duplicated exons and in background (all of the exons in genome) are compared. The comparison showed that the percentages of AS in duplicated exons and in ordinary exons were almost the same (Table S1). However, when using the exons emerged from repeat elements as control, the percentage of AS in repeat-associated exons is much higher, which is consistent with previous study [23]. This difference indicates that AS after tandem exon
duplication is not as prevalent as AS arising from repeat element (Alu exonization). To confirm this observation, we also checked the data in mouse. The result is consistent with human (Table S1).