Biological processes have evolved to decompose recalcitrant biomass not only recruiting a variety of related proteins but also maintaining biomass-specific protein repertoire. Recent advances in computational genomics and proteomics have identified potential protein families deconstructing biomass for bioenergy production. The major obstacle to use them is a complication in predicting correct molecular function, which is often caused by domain multiplicity occurring in protein families. In order to overcome this barrier, we retrieved protein families related to biomass deconstruction from carbohydrate active enzyme database (CAZY) and parsed them into domains. We then constructed a protein domain network from which we deduced interactions and relationships among domains. Based on our analysis, we suggest new classification system for proteins related to biomass deconstruction and annotation procedure for their contextual molecular functions. In specific, our domain-based approach is advantageous for functional prediction of proteins because a protein domain behaves as a functional unit or module during evolutionary history. The algorithm is semi-automatic and we designated the system as protein domains for biomass deconstruction (PDBD) database. The new classification system assists to discover novel protein domains and thus provides a design principle for novel catalysts in biomass deconstruction process for bioenergy production.