Multidomain annotation of carbohydrate-active enzymes beyond CAZy domains with GeneHunt2.

Abstract

Carbohydrate-active enzymes (CAZymes) are central to carbohydrate metabolism, yet their functional annotation is typically restricted to catalytic CAZy domains, overlooking the broader multidomain architectures in which these domains operate. Here, I present GeneHunt2, a scalable framework for multidomain annotation of CAZymes that integrates curated HMM profiles from dbCAN and Pfam into a unified, deduplicated database, enabling systematic identification of both CAZy and non-CAZy domains. After confirming the robust recovery of CAZy domain assignments using GeneHunt2, I investigated the detailed multidomain architecture of over 3.75 million CAZyme sequences: more than 40% were multidomain, and non-CAZy partner domains constituted a substantial fraction of detected partner domains. Using a quantitative framework that combines co-occurrence enrichment, domain adjacency, positional bias, and partner-specificity scoring, I next distinguished family-specific modules, auxiliary domains, and promiscuous partners. This approach recapitulates known CAZyme-CBM relationships and extends beyond CAZy definitions by identifying numerous Pfam domains including many domains of unknown function (DUFs) that are specifically and non-randomly associated with particular CAZy families. By enabling reproducible, multidomain-aware annotation, GeneHunt2 facilitates data-driven hypotheses about poorly characterized domains and widens the functional interpretation of carbohydrate-active proteins beyond their catalytic cores.