Alternative splicing is a sophisticated nuclear process that regulates gene expression [1] and is an important mean of enhancing protein diversity [2]. Alternatively spliced variants play important roles in development and defects in mRNA splicing are causes of many human diseases [3, 4, 5]. Information about alternatively spliced variants is mainly derived from primary transcripts and little knowledge exists regarding structures. Here we carry out a structural genomics analysis of known alternative splicing events and show that threading is a valid approach to model structures of alternatively spliced variants. We collect 3-D structures of proteins with know translated splicing products from PDB and further expand the dataset with high quality models generated with threading approach. Our analysis shows that splicing events have a strong preference for non-regular secondary structure elements and tend to avoid buried residues. Those observations suggest evolutionary constrains exist for locations of splicing in the context of 3-D environment. We then show that majority of substitutions in splicing events share high structural similarity and splicing events also tend to remove entire domain and avoid exposing hydrophobic cores when part of a domain was removed. Those observations support the notion that majority of splicing isoforms adopt same fold as full-length protein despite sequence substitutions and deletions.