The work sits in the mid-2000s effort to catalog alternative splicing, the mechanism by which one gene yields multiple transcripts (and proteins) by including or excluding different exons. HExDB's stated motivation is concretely practical rather than purely descriptive: it was built to help design primers for amplifying exons from cDNA, and to understand how alternative splicing alters open reading frames — that is, how splice choices change the protein a gene encodes.
Methodologically, the database integrates three independent data sources of differing provenance: AltExtron, a computationally predicted exon database; Ensembl cDNA annotation; and a recently published Affymetrix genome tiling array. Combining a predicted set, a curated annotation set, and a tiling-array signal was the paper's approach to building a fuller picture of human exon composition than any single source would give.
Honest context: this is an early-career database paper, and the available source is the journal's article page rather than a fully extractable methods section, so this entry reports what that page states — purpose, data sources, and broad scope — and deliberately stops short of detailed pipeline or validation claims. It is most relevant biographically: it marks the bioinformatics origin (KAIST) of a researcher whose later work moved decisively into behavioral medicine and control engineering for digital health.