Welcome to AFP-R, a resource for Antifreeze proteins

Antifreeze proteins (AFPs) are a special class of proteins that confer freeze tolerance to organisms inhabiting frigid habitats. Their unique structural features enable binding to ice crystals, which depresses the freezing point of bodily fluids via thermal hysteresis (TH) activity or suppresses ice recrystallization (IRI). The absence of a specialized, comprehensive resource for AFPs impedes systematic investigations into their functional mechanisms as well as related applications.

AFP-R is an integrated resource encompassing two core components: AFP-DB, a database housing manually curated antifreeze proteins mined from 256 publications (by August 20, 2025), all with experimentally validated antifreeze activity; and AFP-Predictor – a predictive tool trained on datasets sourced from AFP-DB. AFP-DB archives comprehensive information of antifreeze proteins, including amino acid sequences, three-dimensional structures, modifications, source organisms, raw experimental measurements of antifreeze activity, and associated annotations. In this version, antifreeze glycoproteins are not included. The database currently holds 186 primary entries, 607 subentries, and 1,444 experimental records documenting antifreeze activity assays. AFP-Predictor is built on the protein language model ESM2, and only protein sequences are required as input. Users can use the AFP-R online freely.