A large number of modular domains that exhibit specific lipid-binding properties are present in many membrane proteins involved in trafficking and signal transduction. These membrane lipid-binding domains are present in eukaryotic peripheral membrane and transmembrane proteins. During the last decade computational methods that identify such proteins have been developed, but the current predictors identify only a fraction of these proteins.
Here we report a profile Hidden Markov Model based method capable of predicting membrane binding proteins (MBPs). MBPpred can identify MBPs that contain all the membrane binding domains that have been described to date and furthermore can separate proteins based on their relative position in the membrane plane.
After an extensive literature search we were able to identify 18 domains (Annexin, ANTH, BAR, C1, C2, ENTH, Discoidin, FERM, FYVE, Gla, GRAM, IMD, KA1, PH, PX, Tubby, PTB, GOLPH3) that are associated with binding to membrane lipids. For each of these domains we isolated at least one characteristic pHMM from Pfam. Subsequently, we created a pHMM library containing the 40 pHHMs that were isolated from Pfam. The proteins that are detected from our method, are further classified - based on their interaction with the membrane plane - into transmembrane and peripheral membrane proteins with the use of an algorithm, developed in our lab, PredClass.