Background/Motivation:

The dearth of structural information on alpha helical membrane protein (MPs) has hindered thus far the development of reliable knowledge –based potentials that can be used for automatic prediction of trans-membrane (TM) protein structure. While algorithm for identification of TM segments is available, modelling of the domains of alpha helical MPs involves assembling the segments into a bundle. This requires the correct assignment of the buried and lipid-exposed faces of the TM domains. 

Results: In a cross validated test on single sequences, our trans-membrane MM, correctly predicts the entire topology for 77% of the sequences in a standard dataset of 86 proteins with supervised topology. These results compare favorably with existing methods. 

Source Code: Matlab

Conclusion/Implementation: Here discriminant data mining approach was used to predict the location and orientation of alpha helices in membrane-spanning proteins. It is based on a first order Markov model (MM) with an architecture that corresponds closely to the biological systems. The model is enriched with three types of states for the loop on the cytoplasmic side (outer loop), loop for the non-cytoplasmic side (inner side), and trans-membrane part. The closed association between the biological and Markov states allows us to infer which part of the model architecture are important to capture the information which encodes the membrane topology, and gain a better understanding of the mechanism and constraints involved. Predictor Model was established by various  Markov decoder , and assignment of the membrane helix boundaries was apparent.