Recognizing Antifreeze Proteins with Machine Learning

Steffen Graether (MCB)
Dan Ashlock (Math and Stats)

Many cold-blooded organisms (bacteria, fungi, plants, insects and fish) survive the cold by producing antifreeze proteins (AFPs). AFPs function by preventing the growth of ice at sub-zero temperatures. A central dogma of biochemistry is that the structure determines the function of a protein, and vice versa. AFPs are an extreme example of a group of proteins that break this rule – despite a convergent function (inhibiting ice growth), diverse organisms produce AFPs with diverse sequences and structures. This project applies various machine learning techniques (such as side effect machines) developed for another family of cold-stress proteins, (dehydrins) to known antifreeze proteins in order to locate a family of features that are able to characterize AFPs as a class and could be used to identify novel AFPs.