AFP-R: An Open Resource Dedicated to Antifreeze Proteins

Antifreeze proteins (AFPs) possess the ability to lower the freezing point of substances through thermal hysteresis activity and/or ice recrystallization inhibition. This unique property allows AFPs to protect organisms from freezing damage in cold environments. AFPs have significant potential applications in fields such as biomedicine and agriculture, including tissue-organ cryopreservation and the development of frost-resistant crops.

However, the absence of a comprehensive resource dedicated to AFPs has hindered the advancement of our understanding of their functional mechanisms and hindered their broader applications. In response to this need, researchers have developed AFP-R, an online resource that comprises two key components: AFP-DB and AFP-Predictor. AFP-DB is a comprehensive database containing manually curated proteins with experimentally validated antifreeze activity, derived from published literature.

AFP-Predictor, on the other hand, is a sequence-based machine-learning model designed to identify AFPs. AFP-DB stores a wealth of diverse AFP-related information, including sequences, structures, post-translational modifications, taxonomy, and annotations of antifreeze-activity experimental assays. Currently, AFP-DB holds a total of 186 entries, 607 sub-entries, and 1444 experimental records.

AFP-Predictor, a cutting-edge AFP-identification algorithm, is built upon the protein language model ESM2 (Evolutionary Scale Modeling2). This model has been trained on data from AFP-DB and has demonstrated superior performance compared to several existing AFP identification models. By providing a valuable resource for the systematic study of the mechanisms underlying AFP antifreeze activity, AFP-R is expected to facilitate further research and applications of AFPs in various domains.