Protein-Membrane Interactions

A large number of cytosolic proteins bind reversibly to the membrane in order to perform their function. This reduces a three-dimensional search for a binding partner to a restricted two-dimensional search; this search is restricted even further to specific lipids following specific binding interactions. A number of peripheral membrane-binding proteins have one or more domains specialized in binding lipids; these targeting domains including: C1, C2, PH, FYVE, PX, ENTH, ANTH, BAR, FERM and tubby domains. Likewise, there are peripheral membrane-binding proteins that utilize a portion of the molecular surface rather than a separte domain. Currently, these proteins can be identified in painstaking experimental and translocation studies. Indeed, extending these experiments to the genome scale is not currently possible and traditional sequence analysis techniques cannot transfer sequence similarity to function. Moreover, structural similarity can only transfer function roughly two-thirds of the time.

Protein Function Prediction

In our work, we developed an automated prediction protocol that utilizes structure-based and sequence-based features to identify peripheral membrane binding proteins. In more recent work, we explore the performance of a number of classifiers over this same dataset and features; while SVM is a robust classifiers, it is not known beforehand which classifier works best. Moreover, we analyze the features important to membrane-binding in the context of the model [4,6].