which was done using the OWEN program (36). In all cases, our alignments contained putative exons presented in GenBank. We masked sequences using RepeatMasker (http://humangen.med.ub.es/tools/RepeatMasker.html), because numerous low-complexity and repetitive regions in long intronic sequences obscured the pattern of orthology. All predicted primate transcripts were analyzed by GENSCAN (http://gnomic.stanford.edu/ http://CCR-081.mit.edu/GENSCAN.html). Modified UNCOVER (42) approaches for the recognition of unknown conserved alternatively spliced exons were used for extensive search of regulatory sites and splicing signals. Two alternative approaches for treating gaps (indels) in pairwise sequence alignments were employed: (i) whenever a gap was introduced in one of the sequences, the respective position in another sequence was treated as a non-conserved one (a mismatch) and (ii) positions containing one or more gaps were excluded from the analysis (40). Evolutionary rates at non-synonymous (Ka) and synonymous (Ks) sites were calculated for coding exons using program PALM (69). Evolutionary rates at non-coding exons (Ku) and for intronic sequences (Ki) were calculated using Kimura’s two-parameter model (70). We performed a search for regulatory sites and splicing signals using the TRANSFAC database (http://www.biobaseinternational.com/pages/index.php? id=transfacdatabases). Rodent and primate 5′-UTRs were computationally folded by program Afold (71) to predict potential IRES stable structures. The predicted minimum free secondary
